Solar System Physics for Exoplanet Research

Mercury, the closest planet to the Sun, is a shattered husk of a world. In terms of its uncompressed bulk density, it is the densest of the planets²⁵—the result of a composition that is dominated by metals, rather than silicates or volatiles. From the Earth, Mercury's surface features remain elusive—a result of the planet's small size, and the challenges involved in observing it (since it always remains with some 28° of the Sun in the sky). It was long thought to be trapped in 1:1 mean-motion resonance with the Sun—spinning once on its axis for every orbit—a theory finally that was finally disproved as a result of radar observations of the planet, carried out in 1965 (e.g., McGovern 1965; Pettengill & Dyce 1965). Rather than spinning once per orbit, those observations revealed that Mercury instead spins on its axis once every 59 days, while it orbits the Sun in just under 88 days. As a result, the planet is trapped in a 3:2 spin–orbit resonance—meaning that it completes three full rotations in the time it takes to orbit the Sun twice (e.g., Liu & O'Keefe 1965). No other object is known to be trapped in such a resonance—and it seems likely that the resonance nature of the planet's rotation is maintained, in part, as a result of the planet's eccentric orbit. Were Mercury's orbit circular, it seems likely that the tidal interaction between it and the Sun would continue to slow its rotation, driving it toward an eventual 1:1 spin–orbit resonance. However, as Mercury's orbit is eccentric, the strength of the tides raised upon Mercury by the Sun vary markedly through the course of each orbit—strongest at perihelion, and weakest at aphelion.

The result of Mercury being trapped in this spin–orbit resonance is that the Solar day on Mercury's surface (i.e., the time between two consecutive sunrises for most locations on the planet) is twice the length of Mercury's year. More peculiar still, however, is the combination of Mercury's spin and its eccentric orbit. As the velocity of Mercury in its orbit around the Sun reaches its peak, at perihelion, the easterly motion of the Sun across the sky that results from the planet's motion exceeds the westerly motion that results from Mercury's constant rotation. As a result, the motion of the Sun across Mercury's sky appears to slow as the planet approaches perihelion, then reverses through the perihelion passage (with the Sun moving from west to east). Finally, as Mercury recedes from perihelion, the situation reverses, and the Sun resumes its normal east-to-west motion across the sky. As a result, from some locations on Mercury's surface, the Sun would appear to rise in the east in the morning, slow down, reverse direction, and set once again, before rising a second time and moving across the sky as normal! At other locations, the same kind of behavior would be seen at sunset—the Sun would sink below the western horizon, before rising again, moving west-to-east. It would then slow, turn, and resume its motion, setting once again and ushering in a lengthy 88 day night!

Another result of this peculiar rotation is that there are two "heat poles" on the surface of Mercury—locations where, every second orbit, the Sun reverses direction while close to the zenith. Not only is the Sun overhead (or close to) at perihelion—hence maximizing the instantaneous flux falling on the surface—but it remains there for an extended period of time, thanks to the retrograde motion induced by the speed of the planet's perihelion passage. A huge impact feature, one of the largest impact basins in the solar system, lies close to one of these two "heat poles," with its name (Caloris Planitia, or the Caloris basin; e.g., Gault et al. 1977) intended to invoke the fact it is likely to be one of the hottest locations on the planet, at local noon. The basin shows evidence of ancient volcanism that dates from some time after the basin's formation (e.g., Thomas et al. 2014), indicating that Mercury remained volcanically active until less than a billion years ago.

For many years, it was believed Mercury had no atmosphere—but the first spacecraft to visit Mercury, Mariner 10, revealed that the planet has an extremely tenuous exosphere (e.g., Broadfoot et al. 1974, 1976)—a result confirmed by the more recent Messenger spacecraft (e.g., Solomon et al. 2008; Zurbuchen et al. 2008). That exosphere is not considered bound to Mercury—rather, it is continually generated and denuded as a result of the interaction between the Solar wind and the planet's surface, which leads to a continual erosion and ejection of ionized material from that surface. As such, Mercury's tenuous exosphere consists primarily of a mix of Solar helium, and native sodium, potassium and oxygen ions, with other ionized species forming smaller components—as detailed in Zurbuchen et al. (2008). Interestingly, there is evidence that the impact of meteoroids upon Mercury's surface plays an important role in the generation of its exosphere (e.g., Killen & Hahn 2015, who suggest that passages of the planet through the Taurid meteor stream could lead to annual enhancements in the amount of Calcium found in Mercury's exosphere, and Pokorný et al. 2017, who invoke the seasonal variation in the impact flux at Mercury as a potential origin for the observed dawn-dusk asymmetry in the planet's exosphere).

The Mariner 10 spacecraft gave us our first close-up look at the solar system's innermost planet. In addition to revealing Mercury's tenuous exosphere, it also revealed that Mercury has a significant intrinsic magnetic field, enough to generate a magnetosphere and bow shock (e.g., Ness et al. 1974, 1975), which might indicate the presence of an internal dynamo, and hence that Mercury's core is still in the process of freezing (e.g., Stevenson et al. 1983). The structure of that magnetic field was revealed in more detail by the MESSENGER spacecraft, which made three flybys of the innermost planet in 2008 and 2009, before moving into orbit around the planet in 2011 March (e.g., Solomon et al. 2007, 2008). The spacecraft revealed that Mercury's magnetic field is tilted by ∼5° from the planet's rotation axis (e.g., Anderson et al. 2008, 2010), is approximately 1% the strength of the Earth's magnetic field (at the planet's surface) and is centered on a point that is displaced by just over 400 km to the north of the planet's center (Alexeev et al. 2010). For a detailed review of our knowledge of Mercury's magnetic field following the MESSENGER mission, we direct the interested reader to Anderson et al. (2010).

Mariner 10 also provided our first images of Mercury's surface, and first measurement of the planet's mass and density. Through those flybys, the spacecraft imaged around 45% of the planet's surface, revealing a scarred, cratered world, similar in appearance to the Moon (e.g., Murray et al. 1974). In addition to the abundant impact features of all sizes, more recent observations carried out by MESSENGER have revealed that some ∼27% of Mercury's surface is covered by smooth planes, most of which are thought to be volcanic in origin (e.g., Denevi et al. 2013). In the region surrounding the Caloris basin, MESSENGER revealed a region of hummocky plains, covering approximately 2% of the planet's surface, which might be the result of eject from the impact that formed the basin (Denevi et al. 2013). The imagery returned of Mercury also revealed unusual thrust fault features, known as "lobate scarps," whose origins likely reflect the contraction of the planet as it cooled, following its formation (e.g., Solomon 1977; Watters et al. 1998), with a potential contribution coming from changes in the planet's oblateness as its spin slowed to the current ∼59 day period (e.g., Melosh 1977).

Based on the observations made by Mariner 10, researchers were able to determine that Mercury is differentiated, with a large iron–nickel core that is markedly larger and more massive (as a fraction of the planet's mass) than the cores of the other terrestrial planets (e.g., Murray et al. 1974; Siegfried & Solomon 1974; Gault et al. 1977). These results were refined considerably as a result of the MESSENGER mission—which enabled the construction of a detailed model of Mercury's gravity field (Smith et al. 2012). Those observations revealed that the planet's crust varies in thickness as a function of latitude—being thickest near the equator, and thinnest near the poles. The distribution of mass in Mercury's interior was found to be consistent with the planet having an iron-rich liquid outer core—a result that ties in nicely with the idea that Mercury's magnetic field is generated by an interior dynamo, in much the same way as that of the Earth. As we discuss later (in Section 4.1.1), Mercury's anomalously massive core, and high bulk density, are key pieces of evidence that suggest the planet was once victim to a celestial hit-and-run collision, which shattered the proto-Mercury, and stripped it of much of its silicaceous mantle material (e.g., Cameron et al. 1988; Benz et al. 2007).

For a detailed review of our knowledge of Mercury prior to the MESSENGER mission, we direct the interested reader to the book "Mercury," published as part of the Space Sciences Series of ISSI (Balogh et al. 2008), while Rothery (2015) provides a detailed overview of our understanding of the planet, following the new data delivered by MESSENGER. In the coming years, our knowledge and understanding of the solar system's smallest planet will likely be revolutionized once again, with the arrival of the BepiColombo mission (e.g., Benkhoff et al. 2010). Launched in 2018 October, BepiColombo is following a convoluted path to Mercury—and is scheduled to move into orbit around the iron planet in 2025 December. Before this, the spacecraft will use a mixture of gravitational assists (through flybys of the Earth, Venus, and Mercury itself) and Solar-electrical propulsion to modify its orbit, and gradually reduce its encounter velocity, with respect to Mercury, to one sufficiently slow as to effect the orbit insertion. The spacecraft will then split into two components—the Mercury Planetary Orbiter, and the Mercury Magnetospheric Orbiter (Mio), which will observe Mercury until at least mid-2027. For more details on the goals of the BepiColombo mission, we direct the interested reader to Benkhoff et al. (2010).

Venus is the second planet from the Sun, and one of the solar system planets known since antiquity. It is typically the third brightest object visible from Earth, after the Sun and Moon, and is frequently visible soon after sunset or before sunrise as the "evening star" or "morning star." The apparent brightness of Venus largely stems from the combination of its proximity to the Sun and the Earth, the reflecting/scattering properties of its atmosphere, and its size.

Due to its similarity in size to the Earth (0.95 Earth radii), Venus is frequently referred to as our planet's "twin" or "sibling." Furthermore, the mass of Venus is 82% of that of the Earth, which results in the planet having a similar bulk density to our own (5.204 g cm⁻³ compared to Earth's 5.514 g cm⁻³). It is therefore considered reasonable to assume that Venus has both a comparable bulk composition, and underwent a similar formation process to the Earth (e.g., Zharkov 1983). Up until as recently as the early 1960s, the Venusian surface was theorized to have a range of temperate surface environments, including steamy jungle landscapes. Ground-based microwave observations (Mayer et al. 1958) combined with data from the NASA Mariner (e.g., Barath et al. 1964) and Soviet Venera space programs (e.g., Avduevskij et al. 1971; Keldysh 1977) finally revealed the extent to which the surface of our sister planet diverged from a temperate Earth-like environment (e.g., Sagan 1962; Ronca & Green 1970; Marov 1978).

The first suggestions that Venus might not be a true twin to the Earth came in the early part of the 20th Century. Pioneering spectroscopic observations by Slipher (1903) strongly suggested that the planet's rotation must be far slower than that of the Earth. Indeed, his observations found no evidence of any rotation—leading Slipher to conclude "A glance at the table will show that the errors of observation were small and that there is no evidence of a short rotation period for the planet. A rotation period of twenty-four hours would incline the planetary lines one-third of a degree, a quantity quite large in comparison with the errors of observation." Confirmation of Venus' incredibly slow rotation came with the advent of radar observations, in the 1960s (Goldstein & Carpenter 1963), which revealed both that the planet rotates incredibly slowly (with a period of slightly more than 243 days), and that that rotation is retrograde in nature (i.e., occurs in the opposite direction to its orbital motion).²⁶ As a result, Venus is the planet with the slowest rotation (both in terms of rotational velocity at the planet's equator, and rotation period) and the greatest axial tilt (with a value of around 177°). As we discuss in Section 4.1.1, a number of explanations have been put forward to explain Venus' peculiar rotation—ranging from torques exerted on the planet by its thick atmosphere (e.g., Dobrovolskis & Ingersoll 1980) to the chaotic evolution of the spin under the gravitational influence of the other planets (Correia & Laskar 2001), or even the effects of a giant collision, early in the planet's youth (Dormand & McCue 1987; McCue & Dormand 1993). Regardless of the true origin of the planet's slow spin, it places the planet in an unusual situation—unlike the Earth and Mars, whose atmospheres rotate along with their host planets, the atmosphere of Venus instead super-rotates (aside from the layers adjacent to the planet's surface)—with the upper layers of the atmosphere rotating with a period of approximately four days (e.g., Schubert et al. 1980).

The atmosphere of Venus is dominated by CO₂ (96.5%), with the remainder consisting mostly of N₂ (3.5%). Other trace atmospheric constituents include SO₂ and clouds of H₂SO₄ that extend from 25 to 50 km altitudes above the surface. The thickness of the atmosphere creates enormous pressure at the planet's surface, with an average surface pressure of 92 bars, equivalent to the pressure at an ocean depth of almost 1 km. The scattering properties of the thick atmosphere produce a relatively high Bond albedo of 0.76 (Haus et al. 2016). Venus receives 1.91 times the incident flux received at Earth, but less than 3% of that incident flux penetrates to the Venusian surface. However, the Venusian climate is locked in a post-runaway greenhouse state, with an average surface temperature of 735 K (Walker 1975).

The present state of the Venusian atmosphere is presumed to have been caused through the planet's passage through a threshold whereby the insolation exceeded the outgoing thermal radiation limit for a moist atmosphere (Komabayasi 1967; Ingersoll 1969; Nakajima et al. 1992; Goldblatt & Watson 2012; Goldblatt et al. 2013). Moreover, it is possible that Venus may have retained substantial surface liquid water until as recently as a billion years ago (Way et al. 2016), although it has also been argued that the planet's water may never have condensed after its accretion (Hamano et al. 2013). The evidence for a previously habitable Venus is critically important both for the general evolution of terrestrial planets and the studies of Venus that are applicable to exoplanets (Kane et al. 2018, 2019b).

The interior and geology of Venus retain numerous outstanding questions, including the size and state of the core, refractory elemental abundances, seismic activity and density variations with depth, and convection within the mantle with relation to current surface geology. Pioneer Venus produced one of the first complete topographical maps of Venus via radar mapping (Pettengill et al. 1980), shown in Figure 7. The map reveals the complexity of the surface, including the dominant highlands of Ishtar Terra in the northern hemisphere and Aphrodite Terra along the equator (Ivanov & Head 2011).

Zoom In Zoom Out Reset image size

Images from the JAXA Akatsuki spacecraft show strong evidence for stationary waves in the upper atmosphere, shown in Figure 8, where the wave features are centered above the Aphrodite Terra highland (Fukuhara et al. 2017b). Such atmospheric phenomena are explained by stationary gravity waves caused by deep atmosphere wounds over the highland regions, and emphasize the importance of interactions between planetary atmospheres and surface topography. Indeed, topographical maps of the Venusian surface yield the appearance of analogues to continents and ocean basins. However, unlike Earth, the surface of Venus does not represent a broad distribution of ages (Head 2014).

Zoom In Zoom Out Reset image size

Analyses of surface crater counts have shown that the average age of the surface is ∼750 Ma (Schaber et al. 1992; McKinnon et al. 1997), which is less than 20% of the total age of the planet. The precise mechanisms through which resurfacing scenarios could have occurred is still a matter of some debate (Bougher et al. 1997) and include catastrophic overturn (Parmentier & Hess 1992) and mantle convection mechanism changes (Herrick 1994) as possible explanations. Answering such questions regarding the interior and atmosphere will shed new light on the evolution of habitable terrestrial planets.

The Earth is, of course, by far the best studied of the planets, and, as a result, is the one whose nature is the best understood. As the cradle of life, our planet remains the only known inhabited world—a fact that helps to drive the global community's fascination with the search for another "Earth-like" world—a phrase which means different things to different people. While it is beyond the scope of this review to describe all of the contemporary research into our terrestrial home, we here provide a general overview of our planet's primary physical characteristics, and describe those that mark the Earth as distinct from its neighbors in the solar system.

The Earth is the largest and most massive of the solar system's four terrestrial planets. As our home, it is by far the best studied of all of the solar system's planets—although thanks to our ongoing exploration using robotic spacecraft, it is fair to say that we have better imagery of the surfaces of the Moon or Mars than we do of the bottom of Earth's oceans.

The Earth is accompanied by a single natural satellite—the Moon—whose unusually large size and peculiar composition hint at a violent origin (as we discuss in more detail in Section 4.1.1). The Moon is gradually receding from the Earth, as a result of their mutual tidal interaction—a process that is also slowing the Earth's spin, as angular momentum is transferred between the two bodies. As a result, in the past the Earth–Moon distance was markedly smaller, and our planet's spin faster than that we observe today. In a recent study, de Winter et al. (2020) used core samples taken through a fossilized bivalve (torreites sanchezi) that lived 70 million years ago to show that the year, at that time, was 372 days long. Since it is unreasonable to assume that the Earth's orbital semimajor axis was markedly larger at that time, those results suggest that the length of the Solar day on Earth at the time was 23 hr 31 minutes—some 29 minutes shorter than that we observe today. As well as offering a fascinating insight into life in the late Cretaceous, that work offers a promising new technique that could allow researchers to accurately track the evolution of the Earth's spin over time. Such work is particularly important, since it seems likely that, shortly after the formation of the Moon, the Earth may have spun with a period of just a few hours (e.g., Canup 2012; Ćuk & Stewart 2012; Wisdom & Tian 2015).

Unlike Venus and Mars, the Earth has an appreciable magnetic field, one that is approximately one hundred times stronger than that of Mercury. Earth's strong magnetic field is thought to be produced by a deep internal dynamo generated within the liquid outer core of our planet (e.g., Buffett 2000). This magnetic field produces a large magnetosphere, which acts to deflect the majority of the Solar wind. Those charged particles that succeed in penetrating the Earth's magnetic field follow the field lines down to hit Earth's atmosphere in its polar regions, driving continual aurora at high magnetic latitudes. At times of particularly energetic Solar activity, those aurorae are driven equator-wards, with the most energetic events causing aurorae that can be seen close to the equator. It has been argued that the presence of Earth's strong magnetic field has played an important role in ensuring the ongoing habitability of our planet—with the atmospheric degradation experienced by Mars usually held up as an example of what can happen to the atmosphere of a terrestrial planet without the twin protections of a strong magnetic shield, and outgassing and recycling of the atmosphere driven by plate tectonics. For an overview of the impact of plate tectonics and the Earth's magnetic field on our planet's habitability (along with the various other factors that may have combined to make our planet a suitable host for life), we direct the interested reader to Horner & Jones (2010a), and references therein.

The orientation of Earth's magnetic field is far from fixed—with the location of the magnetic north pole on Earth's surface having drifted markedly over the past three centuries. In the mid-1800s, the northern magnetic pole was located more than twenty degrees from the geographical north pole. In recent years, the orientation of Earth's magnetic axis has been shifting rapidly, with the spin and magnetic poles currently separated by less than four degrees (Chulliat et al. 2019). On longer timescales, there is evidence that Earth's magnetic field can vary markedly in strength, and its orientation can reverse. Such reversals have been demonstrated in numerical models of the Earth's dynamo (Kuang & Bloxham 1997) and, indeed, the periodic reversals in Earth's magnetic field provide an elegant piece of evidence for the ongoing process of plate tectonics on Earth. When iron-bearing rock forms, it freezes in place a record of the orientation of Earth's magnetic field. It was observed that changes in the remnant magnetism in rocks across regions signified that not only had the Earth's magnetism changed during geological time, but also gave large clues as to how the continents had moved relative to each other (Runcorn 1956).

The dynamics of the Earth's surface reflects its interior structure, the deepest layers of which having been probed by seismic investigations. Indeed, the Earth is almost unique among the planets in that it has been subject to a detailed seismic study of its interior, although the NASA InSight mission, currently on the surface of Mars, is aiming to remedy this, with the goal of using seismic observations to investigate the red planet's interior, as well as the flux of impacts it experiences (e.g., Banerdt et al. 2012; Dauber et al. 2018). In the case of the Earth, our ability to continually monitor and study seismic activity has allowed the interior of our planet to be studied in remarkable detail. By monitoring the propagation of seismic waves generated by earthquakes through our planet we have been able to discern that the Earth has two-component core, a liquid outer core (Oldham 1906) and a solid inner core (Lehmann 1936). Surrounding this is a plastically deforming mantle that comprises most of the volume of the planet, and itself can be divided into several layers that are characterized by changes in mineralogy (e.g., Tackley et al. 1994). It is the heat flow within the mantle, through convection, that is the main driving force for movement of Earth's crust, in the form of plate tectonics—a process that is notably absent on the surfaces of the other terrestrial planets. Interestingly, it has been argued that the Earth can only support plate tectonics as a result of abundant water trapped within the mantle. Indeed, the lack of global plate tectonics on Venus, whose size, mass, and composition are similar to those of the Earth, has often been attributed to the lack of water on the planet's scorching surface (e.g., Nimmo & McKenzie 1998; O'Neill et al. 2007). Even water, however, might not have been enough on its own to allow Earth to begin and sustain plate tectonics—with a recent study suggesting that modern tectonic processes might well have been initially triggered by the impacts that bombarded the Earth in the billion years after its formation (e.g., O'Neill et al. 2019).

The Earth's atmosphere is deep and complex. By volume, when dry, it comprises approximately 78% nitrogen, 21% oxygen, and 0.93% argon, 0.04% carbon dioxide,²⁷ and traces of a wide variety of other elements and compounds. In the lower atmosphere, water vapor provides an additional component, though the amount of H₂O can vary quite dramatically, from almost none in deserts and at times of particularly low humidity, to being super-saturated. The Earth's atmosphere is noticeably stratified, and is typically broken down into several major layers. Closest to the ground is the troposphere, within which the great bulk of the planet's weather and clouds can be found. Within the troposphere, in general, the temperature decreases with increasing altitude. The second layer of Earth's atmosphere, the stratosphere, sees this trend reverse, with temperature once again increasing as one rises farther above Earth's surface. This temperature inversion may well have played a major role in helping to ensure that the Earth has not become dehydrated over the aeons since it formed, by serving as a "cold trap." Water vapor which rises too high in the atmosphere will condense out, freezing, effectively forcing it to fall back to lower altitudes. As a result, almost all of our planet's water vapor remains trapped beneath this layer—well below the altitudes at which solar ultraviolet radiation (most of which is blocked by the ozone layer) would dissociate that water vapor, allowing its constituent hydrogen to escape from our planet. Above the stratosphere, the Earth's atmosphere begins to cool once again—a region known as the mesosphere, which extends to an altitude of approximately 80 km above the ground. Then, once again, the temperature of the atmosphere begins to climb, rapidly—a region known as the thermosphere. Beyond the thermosphere lies the exosphere—the region where the Earth's atmosphere gradually thins until it becomes indistinguishable from the interplanetary medium.

Perhaps the most striking feature of the Earth is the presence of liquid water across our planet's surface. Indeed, oceans, rivers, and lakes cover something like 70% of the Earth's surface—a stark contrast to the surfaces of the other terrestrial planets. The origin of that water remains heavily debated—a topic we discuss in more detail in Section 4.4. While it has been suggested that both Mars and Venus were once warm and wet, like the Earth, our planet is the only one of the three whose climate has remained suited to the presence of widespread surface liquid water. Over the course of the Earth's history, our planet's climate has remained remarkably stable—a fact made even more remarkable by the fact that, over the same time period, the Sun's luminosity has increased by an estimated ∼30%. The fact that the Earth's climate was clement at a time when the Sun was just 70% as luminous as it is today led to a conundrum known as the "faint young Sun problem." For many years, the Earth's early warm climate appeared to be at odds with the fact that the young Sun was so faint. For a detailed overview of the "faint young Sun problem," we direct the interested reader to Feulner (2012), and references therein.

Despite its apparent stability on billion years timescales, the climate of the Earth does vary to some degree on shorter timescales. The most famous of these periodic oscillations are known as the Milankovitch cycles—and are the result of the continual perturbation of the Earth by the gravitational influence of the other planets in the solar system. For a more detailed discussion of the Milankovitch cycles and the history of the Earth's climate variability, we direct the interested reader to Horner et al. (2020), and references therein.

Mars, the outermost of the terrestrial planets, is probably the most studied object in the solar system, aside from Earth. It is markedly smaller than the Earth (with an equatorial radius of just 3396 km, compared to Earth's 6378 km), with mass just 0.107 times that of our planet. It is also significantly less dense than the other terrestrial planets, with a bulk density of 3.934 g cm⁻³, which recent work has suggested might be the result of the planet being starved of material during its formation as a result of the migration of Jupiter (e.g., Brasser et al. 2016), as we discuss in section four. Unlike the other planets visible with the unaided eye, which typically appear to have a vaguely yellowish hue, Mars appears a striking red—the result of the iron oxide that dominates the planet's surface.

During the 19th Century, telescope observations of Mars began to reveal a world that displays some similarities to the Earth. Like Earth, Mars has polar caps whose size waxes and wanes with the passing seasons. While Earth's polar caps are dominated by water ice, we now know that those on Mars are a combination of a "bedrock" made of water-ice and carbon dioxide ice, with their expansion and contraction caused primarily by seasonal deposits of carbon dioxide ice—the result of the winter temperatures at the poles being far colder than those on the Earth. Those observations also revealed dark areas on Mars' surface, which some considered to potentially be areas of vegetation (e.g., Hollis 1908), though we now know that they are instead highland regions, and bedrock scoured clear of Mars' ubiquitous reddish dust. Toward the end of the 19th Century, speculation on the possibility that there could be widespread life on Mars—potentially even advanced, technological life—reached fever pitch (e.g., Flammarion 1892; Comstock 1902; Lowell 1908), with speculation fueled by the observation of "canali" (or channels) on the planet's surface. The concept of the Martian Canals was born, primarily through a mistranslation of the Italian "canali" (e.g., Green 1879; Maunder 1888; Lowell 1895; Hamilton 1916). As telescopes improved, however, it gradually became accepted that these "canali" were actually optical illusions/artefacts (e.g., Evans & Maunder 1903)—though debate over the topic continued to rage through to the dawn of the space age (e.g., Webb 1955, and references therein). Nevertheless, through the first half of the twentieth century, it became clear that the modern Mars is far from an ideal location for widespread, complex life (e.g., Tombaugh 1950, who wrote "It appears likely that Mars has always had a thin atmosphere, very little water, and a very dry climate"). The first images returned of Mars from space drove the final nail into that particular coffin—revealing an arid, cratered, cold and desolate world (e.g., Chapman et al. 1968), and confirming the lack of canals on the planet's surface (e.g., Sagan 1975).

Over the past few decades, the idea that Mars might once have supported life, or at least supported conditions suitable for the development of life, has played an important role in directing our ongoing exploration of the red planet. The first spacecraft to successfully visit Mars was the American Mariner 4, which flew rapidly past the planet in 1965 July, returning a small amount of data and imagery (e.g., Fjeldbo et al. 1966; Chapman et al. 1968; Siscoe et al. 1968). The first mission to orbit Mars, the American Mariner 9, in 1971 November (e.g., Hanel et al. 1972; McCauley et al. 1972; Conrath et al. 1973; Sagan et al. 1973), was quickly followed by the first successful soft landing on the surface of the red planet, by the Russian mission Mars 3 (which failed some 20 s after reaching Mars' surface). The Viking missions, in the mid-1970s, were a huge success—both Viking 1 and Viking 2 comprised a Mars orbiter and a lander—with the two landers surviving and operating on the surface of Mars for several years (six for Viking 1, and three for Viking 2). The Viking landers provided our first in situ measurements of Mars surface, yielding a vast amount of useful data (e.g., Hess et al. 1977; Toulmin et al. 1977; Adams et al. 1986, and many others). The landers carried a suite of three biological experiments, designed to search for any evidence of life in the soils at their landing site—the results of which were inconclusive, at best (e.g., Klein et al. 1976; Levin & Straat 1976; Klein 1977, 1978).

After the success of the Viking landers, our exploration of Mars underwent a hiatus of almost two decades, before undergoing a renaissance in the second half of the 1990s. In the 21st Century, Mars has been peppered with a series of landers and rovers, with new missions departing to the red planet with every opposition (with the mean synodic period of Mars being approximately 780 days). The Mars rovers Spirit and Opportunity were a particular highlight of the 21st Century exploration of Mars. The two rovers landed on Mars in 2004 January, for a mission initially scheduled to last just 90 Sols (1 Sol is one Martian day, 24^h 37^m). Both rovers vastly outlived their expected lifetimes—Spirit remained active for more than six years, while Opportunity's journey across Mars was finally brought to an end as a result of a catastrophic loss of power resulting from an intense Martian dust storm on 2018 June 10th, after a remarkable 5352 Sols. In that time, Opportunity covered a distance of more than 45 km—meaning that it holds (by a comfortable distance) the record for the greatest distance traveled over the surface of another world. At the same time, a series of Mars orbiters have returned exquisite images of the planet's surface—with the result that the surface of Mars has been more closely studied than that of the Earth (given the challenges inherent in imaging the bottom of the Earth's oceans).

The result of all that work is that Mars is better studied and characterized than any other object in the solar system (except the Earth). The earliest missions to Mars revealed that the planet has essentially no magnetic field (e.g., Smith et al. 1965), though parts of the planet's crust retain evidence of ancient magnetism (Acuña et al. 1999; Langlais et al. 2004), evidence supported by measurements of the Martian meteorite ALH84001 (Weiss et al. 2002). Based on that evidence, it seems likely that the young Mars had a molten, iron-rich core, which allowed the planet to generate a strong magnetic field. Unlike the Earth, which has maintained its core dynamic until the current day, Mars lost its dynamo, and thence its magnetic field, after a few hundred million years—potentially as a result of the planet cooling more rapidly than the Earth (e.g., Stevenson 2001, 2003; Williams & Nimmo 2004). The surface of Mars shows that, unlike the Earth, the planet does not undergo plate tectonics. Indeed, the lack of plate motion has allowed a cluster of vast shield volcanoes to grow above a hotspot, forming the largest volcanoes in the solar system (the largest of which, Olympus Mons, is almost 26 km tall, as measured from its base, and has a surface area only marginally smaller than that of France; e.g., Mouginis-Mark 2018). For more information on Mars' volcanic history, see Werner (2009), and references therein.

There is abundant evidence that Mars was once warm and wet, with an ocean potentially filling the planet's vast northern lowlands (e.g., Helfer 1990; Clifford & Parker 2001; Carr & Head 2003; Rodrigeuz et al. 2015). This in turn suggests that the planet's atmosphere must once have been much thicker and warmer than that we see today (e.g., Ramirez 2017, and references therein). It seems likely that the lack of a global magnetic field for the bulk of Mars' evolution has allowed Mars' atmosphere to be denuded by the influence of the Solar wind. The lack of plate tectonics on the red planet has also played a role in the loss of its original thick atmosphere, since without plate tectonics, the planet lacks a mechanism by which any gas chemically trapped in the planet's surface rocks could be recycled (e.g., Tomkinson et al. 2013). For a more detailed discussion on the loss of Mars' atmosphere, we direct the interested reader to Horner & Jones (2010a), and references therein. Interestingly, recent radar observations of Mars' south polar cap have revealed that there may exist a permanent reservoir of liquid water on modern Mars—in the form of large lakes buried beneath the ice of Mars' southern polar cap (Orosei et al. 2018). It seems likely that future Mars mission will look to explore this in greater detail, since such deeply buried lakes in Antarctica are known to teem with life (e.g., Christner et al. 2014). If life ever did find a foothold on Mars, then it seems plausible that it could still survive, buried deep beneath the Martian polar ice.

Tied to these ideas of Mars' ancient oceans, it is worth noting that Mars' exhibits a remarkably varied range of surface features—from the vast Valles Marineris (a valley over 4000 km long, 200 km wide, and up to 7 km deep) to the vast volcanoes of the Tharsis bulge. Perhaps the most interesting feature of Mars' global topology, however, is known as the Martian Dichotomy. Aside from the vast impact basin Hellas, Mars' southern hemisphere consists of highlands, which are heavily cratered, and considered to be ancient terrain. The northern hemisphere, by contrast, consists of smooth lowlands, with few if any scars. The Dichotomy is show below, in Figure 9, based on data from the Mars Orbiter Laser Altimeter on Mars Global Surveyor. That dichotomy may well be the scar left behind by a cataclysmic impact on the planet (e.g., Andrews-Hanna et al. 2008), as we discuss in more detail in Section 4.1.1.

Zoom In Zoom Out Reset image size

The robotic exploration of Mars is scheduled to continue over the coming years. Of particular interest is the NASA InSight mission, which landed on Mars in 2018 November. InSight carries a sensitive seismometer, and a thermal probe, which together will provide a wealth of fresh insights on the nature of Mars' interior—from the thickness of the planet's crust, and new information on the radius and density of the planet's core (e.g., Folkner et al. 2018), to the rate at which heat flows outward from the planet's interior (e.g., Spohn et al. 2018), and even new data on the frequency with which Mars experiences impacts from cometary and asteroidal bodies (e.g., Teanby 2015; Dauber et al. 2018).

Early highlights from the InSight mission include the discovery that the local magnetic field around InSight's location is an order of magnitude stronger than that estimated from instruments in orbit around Mars, which is thought to be the result of magnetized rocks buried beneath the surface near the lander, and the finding that that field is time variable, likely as a result of activity high in the planet's atmosphere (e.g., Banerdt et al. 2020; Johnson et al. 2020). Perhaps the highest profile results from InSight to date have come from the SEIS instrument, designed to measure Mars' seismic activity (Lognonné et al. 2019). That instrument has proved hugely successful in detecting Marsquakes, having cataloged several hundred to date. The largest of those quakes occurred in the Cerberus Fossae region, revealing that the faulting observed in that area remains geologically active today (e.g., Witze 2019; Banerdt et al. 2020; Giardini et al. 2020), and it will be fascinating to see to what degree that activity continues in the coming years. For all that Mars is by far the best studied celestial body, it seems certain that the coming decade will deliver a wealth of new surprises, as our ongoing exploration allows us to learn ever more about the red planet.

The gas giant planets Jupiter and Saturn contain most of the planetary mass in solar system (at 318 and 95.2 M_Earth, respectively; see Table 1). Although they are similar in diameter (with Saturn's equatorial diameter being just 17% smaller than that of Jupiter), Jupiter is more than three times the mass of Saturn, and as a result, it is bulk density is markedly higher (1.326 g cm⁻³ versus 0.687 g cm⁻³). In addition to comprising the bulk of the mass of the solar system (aside from the Sun), Jupiter and Saturn also comprise the majority of the angular momentum budget of the entire system—a fact that points to the importance of the giant planets and their evolution on the architecture of planetary systems.

Both planets maintain systems of rings composed of rocky and ice particles of varying size. The presence of rings around Saturn has been known since the seventeenth century (as described in Section 4.1.1)—in part as a result of the planet's axial tilt of almost 27°.²⁸ When they are close to edge on, the rings disappear from the view of all but the largest telescopes—and if Saturn's axial tilt were as low as that of Jupiter (a meagre 313), it seems quite likely that they would have evaded such early telescopic discovery. In contrast to Saturn's magnificent ring system, Jupiter's rings were not discovered until the Voyager 1 spacecraft visited the planet in 1979 March (see Miner et al. 2007, and references therein). Recent observations have revealed that such ring systems may well be transient companions for the giant planets, providing evidence that Saturn's ring may well become entirely depleted as a result of an ongoing process of "ring rain" within the next 300 million years (O'Donoghue et al. 2019). The question of whether Saturn's rings are primordial or transient has implications for the origin of the system, as we discuss in Sections 4.1.1 and 4.1.3.

Both Jupiter and Saturn host large numbers of natural satellites (with, between them, 161 of the 205 planetary satellites discovered to date), some of which display active volcanic (i.e., Io) or cryovolcanic (i.e., Enceladus) activity. Over the past five decades, Jupiter and Saturn have been investigated in exquisite detail by a series of spacecraft: Pioneer 10 and 11 (e.g., Carlson & Judge 1974; Smith et al. 1974; Gehrels et al. 1980; Null et al. 1981), Voyagers 1 and 2 (e.g., Broadfoot et al. 1979; Smith et al. 1979, 1981, 1982), Galileo (e.g., Carlson et al. 1996; Niemann et al. 1998; Gautier et al. 2001; Vasavada & Showman 2005), Ulysses (e.g., Balogh et al. 1992; Stone et al. 1992; Grün et al. 1993), Cassini-Huygens (e.g., Porco et al. 2005; Dougherty et al. 2009; Fletcher et al. 2010; Kanani et al. 2010), New Horizons (e.g., Baines et al. 2007; Gladstone et al. 2007; Reuter et al. 2007; Spencer et al. 2007), and Juno (e.g., Bolton et al. 2017; Connerney et al. 2017; Wahl et al. 2017).

Jupiter and Saturn are composed of mostly hydrogen and helium at a roughly Solar composition ratio, but both planets are enriched in heavy elements relative to Solar (see Atreya et al. 2016, for a discussion). After diatomic hydrogen, methane and ammonia are the two most abundant molecular species in the atmospheres of Jupiter and Saturn. Global cloud layers consisting of ammonia ice, ammonium hydrosulfide (NH₃SH), and water (H₂O) are present in both atmospheres, although at different pressure levels between 0.1 and 10 bars (see e.g., Ragent et al. 1998; West et al. 2009). Jupiter's global clouds are clearly visible when imaged in true (or enhanced color) images, such as that shown in Figure 10 (taken by the Juno spacecraft). The clouds in Jupiter's and Saturn's atmospheres have been used to measure these planets' zonal winds, although the generation of those winds and their connection to deeper layers of the atmosphere or interior is still an area of active research (e.g., Kong et al. 2018; Galanti et al. 2019). Jupiter and Saturn both have warm stratospheres, within which temperature inversions are driven by the deposition of solar radiation. This radiation drives photochemistry in both atmospheres that generates hazes (e.g., Irwin et al. 1998; Fouchet et al. 2009), which can be seen to reflect solar radiation near 2 microns in Saturn (as can be seen in Figure 11, below). Above the region dominated by neutral species, the extended atmospheres of both Jupiter and Saturn include ionospheres and magnetospheres. These magnetospheres are dynamic (e.g., Mitchell et al. 2009; Connerney et al. 2017) and produce auroral emission (visible in Figure 11) caused by interactions with the solar wind as well as material originating on the surfaces, or in the interiors, of the planet's satellites (e.g., Vasavada et al. 1999; Clarke et al. 2009; Mura et al. 2017).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Both Jupiter and Saturn exhibit differential rotation, with their equatorial regions rotating more rapidly than the poles.²⁹ As a result of their rapid rotation (with periods of order ten hours), the two planets are noticeably oblate.³⁰ Within the interiors of Jupiter and Saturn, molecular hydrogen transitions to liquid metallic hydrogen (H⁺) at a pressure of roughly 2 Mbar (e.g., Fortney et al. 2018). The depth at which this transition occurs in differs between the two planets, as a result of the significant difference in their mass. Deeper still, Jupiter and Saturn contain cores of rock and ice. Previous measurements of the nondimensional moments of inertia of the two giant planets have led to the suggestion that Saturn is more centrally condensed than Jupiter (e.g., Helled 2011; Guillot & Gautier 2015). Indeed, recent high-precision measurements of the gravitational moments of Jupiter from the Juno spacecraft have revealed that Jupiter's core is diluted, and potentially extends to a large fraction of the planet's radius (Wahl et al. 2017). Such diffusion between the core and the envelope of giant planets has been predicted by previous high-pressure equation of state experiments (Wilson & Militzer 2012), but there are also suggestions that this dilution of the Jovian core could be evidence that Jupiter was once victim to a giant collision (e.g., Liu et al. 2019; see Section 4.1.1). Jupiter's core is thought to contain between seven and twenty-five Earth masses of heavy elements (Wahl et al. 2017), while Saturn's core is thought to contain between five and twenty M_Earth of such material (Fortney et al. 2018)—findings that offer strong support to their idea that the two planets formed through a process of core accretion.³¹

Both gas giants are still in the process of cooling from their formation, and this thermal radiation can be readily detected with observations at 5 microns (see Figure 11). However, research suggests that the thermal evolutionary histories of Jupiter and Saturn are markedly different. Homogeneous evolutionary models tend to accurately reproduce the current luminosity of Jupiter, but substantially underpredict that of Saturn (see e.g., Fortney & Nettlemann 2010 for a review on this topic). This discrepancy has presented a longstanding, critical issue to the fundamental understanding of the formation and evolution of the entire class of gas giant objects. Helium phase separation and subsequent "rain out" onto Saturn's core is often used to invoke an extra source of energy within Saturn (e.g., Stevenson & Salpeter 1977a, 1977b). Over forty years after the introduction of this theory, results from the Cassini mission seem to have provided verification (Koskinen & Guerlet 2018). However, modeling and observational efforts will likely continue to explore this topic into the future.

A whole generation of space scientists were inspired when Voyager 2 flew by Neptune in 1989 August, sending us the first images of its dark blue disk. Four years earlier, the same spacecraft had flown past Uranus, and still over 30 yr later remains the only spacecraft to have visited our solar system's ice giants. Many of the observations made by Voyager 2 remain to date the best information we have to constrain our understanding of the ice giants. Based on Voyager 2 results, both Uranus and Neptune have been the subjects of books within the University of Arizona's Space Science Series—"Uranus" (Bergstralh et al. 1991) and "Neptune and Triton" (Cruikshank et al. 1995). However, over the last 30 yr, Earth and near-Earth observations have made a significant impact to our understanding of these worlds. Observations from facilities such as the Very Large Telescope, the Atacama Large Millimeter/submillimeter Array (ALMA), and the Hubble Space Telescope (HST), have shown both the ice giants to be much more dynamic worlds than the Voyager 2 data had suggested.

Why the name "Ice Giants"? This moniker has come from the average densities of Uranus and Neptune, which "weigh" in at 1.27 and 1.65 g cm⁻³ respectively. This early observation led to the conclusion that these planets were markedly enriched with the "heavier" elements of oxygen, carbon, sulfur and nitrogen compared to the composition of Jupiter and Saturn, whose overall makeup reflects more closely that of our Sun (as discussed above). The widely accepted view is that these elements will be in the form of ices; with H₂O, CH₄, H₂S and NH₃ combined making up ∼70% of the mass of both Uranus and Neptune, and that it is the physical properties of these materials that will govern the interactions on these planets.

Though Uranus and Neptune are often collectively written about as "the Ice Giants," it should be pointed out that there are some large differences between these worlds. The starkest of these is the negligible heat flux exhibited flowing from Uranus' interior, as observed by Voyager 2, making its upper layers the coldest planetary location within our solar system. Added to this is the 98° obliquity of Uranus, which results in the planet experiencing extreme seasonal change. In contrast, Neptune has very strong self-luminosity (2.61 times more the solar influx; Pearl & Conrath 1991), and has been continually observed to display distinct meteorological features. As a result, it has been suggested that the two planets represent end-members of a "Neptune and sub-Neptune" class of astronomical object (Fulton & Petigura 2018).

Uranus and Neptune both host a suite of satellites, many of which are discussed in subsequent sections. Uranus's larger, regular satellites (including Titania, Oberon, Ariel and Umbriel) have orbits which are coplanar to Uranus' rotation, tilted to the plane of the rest of the solar system.³² Thirteen of these regular satellites have, to date, been discovered orbiting interior to the orbit of Miranda, the innermost of the five moons known to orbit the planet at the dawn of the Space Age. Ten of those satellites were discovered during the Voyager 2 flyby of the planet (Smith et al. 1986; Jacobson 1998), with an eleventh, Perdita, being discovered more than a decade later using data obtained during the flyby (e.g., Karkoschka 2001). The other two new regular satellites (Cupid and Mab) were discovered in 2003 using the HST (Showalter & Lissauer 2003, 2006). In the last three decades, ground-based observations have revealed a secondary satellite system orbiting Uranus, comprising (at the time of writing) nine much smaller irregular satellites largely moving on retrograde orbits (see e.g., Gladman et al. 1998b, 2000; Sheppard et al. 2005).³³

Neptune's satellite system is dominated by its largest moon, Triton, which stands out as it is the only large satellite within our solar system to move on a retrograde orbit about its planet. It is now considered that the most likely explanation for the origin of Triton is that it was once a binary trans-Neptunian object, moving on an orbit not unlike that of the dwarf planet Pluto, and that it was gravitationally captured by Neptune (Agnor & Hamilton 2006) as a result of an encounter with the planet that tore the binary asunder. That capture event is thought to have had catastrophic implications for the rest of Neptune's satellites at the time. We discuss the origin of Triton in more detail in Section 4.1.3.

Both Uranus and Neptune rotate with similar periods, roughly intermediate between those of the gas giants and the Earth. Uranus rotates once every 17 hr, 14 minutes, while Neptune spins slightly faster, with a period of 16 hr, 6 minutes. These similar rotation periods, coupled with the limited available measurements of the two planet's gravity fields, have led to a common interior structure being proposed for both planets. This is based on a three-layer model (Guillot 1999), comprising of an upper layer of ices mixed with molecular hydrogen and helium, that transitions to an icy mantle of water, methane and ammonia—potentially mixed with silicates. The very center of the ice giants is thought to comprise a dense core of silicates and metals, though some have suggested that carbon could condense to diamond under these conditions (Benedetti et al. 1999).

Our abiding image of Uranus is of a pale turquoise gas world devoid of the tumultuous storms that are enduring on the other gas giants. This imagery, taken by Voyager 2 as it flew past Uranus in 1985 is perhaps unrepresentative, as it is of the planets' Northern hemisphere summer. Due to its extreme obliquity, Uranus is subject to extreme seasons, and during a given hemisphere's winter the majority of it will not see the Sun for 20 yr. On the progression to equinox in 2007, a noticeable increase of cloud activity was observed in Uranus' atmosphere (de Pater et al. 2015), and this continues to be monitored in the hope it will reveal much about Uranus' atmospheric dynamics. Neptune, in contrast, has shown constant atmospheric activity since being first observed by Voyager 2 on its 1989 flyby. This activity is mostly observed in the infra-red and at the planet's mid-latitudes, however a bright equatorial storm tracked by professional and amateur observers through 2017–2018 is challenging our emerging understanding of Neptune's atmosphere (Molter et al. 2019).

Both Uranus and Neptune exhibit magnetic fields that are anomalous to those observed elsewhere in the solar system. Unlike the magnetic fields of Earth, Jupiter and Saturn; those of Uranus and Neptune are not dipolar or axially symmetric. It was suggested (Hubbard & MacFarlane 1980) that to generate fields of this nature a type of thin-shell dynamo would need to arise. Calculations have shown that this is possible (Stanley & Bloxham 2004), but for this to occur a convective region in the interior must occur. This is at odds with many of the current interior models of the two planets, which have assumed purely conductive interiors for Uranus and Neptune.

With the great success of the Galileo, Cassini, and Juno missions in exploring the solar system's gas giants, there is a current ground-swell of support for a similar flagship mission to explore Uranus and Neptune. Our current lack of detailed information about our own ice giants is particularly disappointing given the great number of potential "ice-giant" type planets that have been detected orbiting other stars in recent years (as we discuss in Section 5). While those planets orbit at much smaller radii than Uranus and Neptune, the best laboratories for the detailed study of such planets are, at least in the relatively near future, those in our own planetary system. If we are to understand these far away worlds at all we need first to well constrain our own neighbors Uranus and Neptune.

When we look around the solar system today, many objects bear witness to the giant collisions that occurred during the latter stages of planetary formation. From the impact that created the Moon (e.g., Benz et al. 1986; Reufer et al. 2012) to that which has been invoked to explain the capture of Neptune's giant moon Triton (e.g., Goldreich et al. 1989; Woolfson 1999), it seems that cataclysmic impacts and collisions between planet-sized bodies were a regular occurrence as the planets formed, and played a pivotal role in shaping the solar system as we know it.

4.1.1. Giant Impacts and the Planets

4.1.2. Giant Impacts, the Asteroid Belt, and the TNOs

The evidence of the violent collisional history of our solar system is not limited to the planets themselves. Obviously, the larger the object, the larger the collision must be to significantly disrupt it. It is no surprise, then, that when we look at the small body populations of the solar system, examples abound of objects that at some point suffered giant collisions. Indeed, it is likely that those known are only the tip of the iceberg. In this subsection, we present just a few exemplar cases that reveal the violent history of the solar system's small body reservoirs.

The asteroid belt is a collisionally ground reservoir, which is generally thought to contain only a small fraction of the mass it had at the birth of the solar system (as discussed by e.g., Weidenschilling 1977; O'Brien et al. 2007; Morbidelli et al. 2009; Bottke et al. 2015).⁴⁰ Within the belt, tens of asteroid families have been identified, thought to be the results of catastrophic collisions between asteroids (e.g., Marzari et al. 1995a, 1998; Milani et al. 2014; Nesvorný et al. 2015; Cellino et al. 2019). Unlike the giant impacts that are thought to have affected the planets, the collisional grinding of the asteroid belt is an ongoing process, with new families having been created at all points in the solar system's history (e.g., the Karin family, a sub family of the older Koronis family, which is thought to have formed just ∼5.8 Myr ago; Nesvorný et al. 2002). Over time, the members of these families diffuse through the belt, to a greater or lesser extent, depending on a variety of dynamical and non-gravitational effects, to the extent that some of the families that must have been created by the oldest collisions can no longer be detected (e.g., Hanuš et al. 2019). Figure 12 shows the distribution of known collisional families across the asteroid belt.

Zoom In Zoom Out Reset image size

The X-type asteroids in the inner main belt have recently been shown to comprise, among their number, at least two such widely dispersed collisional families (Delbó et al. 2019). The eldest of these, whose largest member may be the asteroid (689) Zita, has a wide distribution of orbital elements, with members spread across the entirety of the inner belt. That family was only identified after the authors considered the correlation between the size of a given asteroid and the rate at which it will migrate through the belt under the influence on non-gravitational forces (e.g., Vokrouhlický et al. 2006, Milani et al. 2014; Spoto et al. 2015; Bolin et al. 2017). The wide distribution of the members of the Zita family indicates that that family may even have formed prior to the early planetary instabilities invoked by the Grand Tack and Nice Models of planetary migration (Delbó et al. 2017). The second family identified in the inner belt by Delbó et al. (2019) is the Athor family, after asteroid (161) Athor, has an estimated age of approximately 3 Gyr age, and while still widely dispersed, clusters more closely in element space than the Zita family, potentially as a result of its formation coming after periods of instability in the outer solar system.

Some of the fragments ejected in these asteroid–asteroid collisions are ejected onto orbits that evolve, under the influence of both non-gravitational forces (see Section 4.9) and distant gravitational perturbations from the planets, to regions of instability in the belt, and are then injected into the inner solar system, to become the near-Earth asteroids (e.g., Morbidelli 1999; Morbidelli et al. 2002a, 2002b; Morbidelli & Vokrouhlický 2003; Bottke et al. 2015; Granvik et al. 2017). Such material forms a significant fraction of the ongoing meteoritic flux at the Earth, and certain classes of meteorites have been strongly linked to specific asteroid families, or even individual asteroids (e.g., the basaltic achondrite meteorites and asteroid 4 Vesta, Binzel & Xu 1993; Asphaug 1997; Drake 2001; Marchi et al. 2012).

It has been proposed (Bottke et al. 2007) that the fragmentation of the parent body of the Baptistina family, around 160 Myr ago, was the source of the object that collided with the Earth 65 Myr ago, creating the Chicxulub impact crater (e.g Hildebrand et al. 1991; Morgan et al. 1997; Dressler et al. 2003), and potentially causing the Cretaceous/Tertiary (K/T) mass extinction (e.g., Alvarez et al. 1980; Smit 1999; Schulte et al. 2010), in which the dinosaurs went extinct. It should be noted, however, that the direct link between the extinction and the impact is still the subject of some debate (e.g., Reddy et al. 2009), with more recent work supporting the idea that the extinction was in fact the result of a double-whammy—the combined effects of an enormous ongoing volcanic extrusion of flood basalts, forming the Deccan Traps in India, and the impact that created the Chicxulub impact crater (e.g., Peterson et al. 2016). It is even possible that the scale of the eruption that formed the Deccan Traps was influenced by the Chicxulub impact, with Renne et al. (2015) finding that as much as 70% of the total extruded material was erupted within 50,000 yr of the impact.

Even though the known trans-Neptunian population is far smaller than that of the asteroid belt (an artifact of its greater distance from the Sun—we remind the reader that the estimated population of the trans-Neptunian region is far greater than that of the asteroid belt; e.g., Durda & Stern 2000; Gomes et al. 2008), it is already known to contain a number of examples of giant collisional processes. The dwarf planet Pluto is accompanied by five satellites (Charon, Hydra, Nyx, Kerberos and Styx; Showalter & Hamilton 2015), the orbits of which are best explained by a giant-impact formation mechanism like that invoked for the Earth–Moon system (e.g., McKinnon 1989; Canup 2005; Stern et al. 2006; Canup 2011)—a theory supported by the more recent observations performed by the New Horizons mission, which visited the dwarf planet in 2015 (e.g., McKinnon et al. 2017; Sekine et al. 2017).⁴¹

Eris, another of the trans-Neptunian region's dwarf planets, has a satellite, Dysnomia, which may also have been produced in this fashion (Greenberg & Barnes 2008), while a number of the smaller trans-Neptunian objects have similar satellite systems which might have formed in this way (e.g., the Orcus–Vanth system, Brown et al. 2010). However, recent observations of both the Orcus–Vanth and Eris–Dysnomia systems have cast some doubt on the origin of their satellites (Brown & Butler 2018). Those observations reveal that the two satellites in question (Vanth and Dysnomia) have low albedos, similar to those observed for other objects of a similar size in the Edgeworth–Kuiper Belt—rather than the high albedos that might be expected had those moons formed through accretion from an icy disk around their parent bodies resulting from a giant collision. As Brown & Butler (2018) note, further study is needed to definitely determine the origin of the satellites orbiting these two large trans-Neptunian objects.

The dwarf planet Haumea has two satellites, thought to have formed in a giant impact (e.g., Brown et al. 2006), and its own ring system (Ortiz et al. 2017; Winter et al. 2019). In addition, it is the largest object in the first detected collisional family in the trans-Neptunian region (e.g., Brown et al. 2007; Ragozzine & Brown 2007; Schlichting & Sari 2009; Leinhardt et al. 2010; Lykawka et al. 2012; Volk & Malhotra 2012; Vilenius et al. 2018; Proudfoot & Ragozzine 2019). It is almost certain that, over the coming years, more collisional families will be found in the trans-Neptunian region (Marcus et al. 2011). However, since the orbital velocities of objects in that region are far smaller than the orbital velocities of the asteroids, the ejecta from any given collision will disperse far more widely through the region (since the ejecta velocities will be a far greater fraction of the orbital velocity).⁴² In addition, as Marcus et al. (2011) note, the slow collisional velocities may well be "insufficient to disrupt the largest (and most visible) bodies"—with collisions on such objects likely resulting in either accretion, or a hit-and-run scenario for large impact angles (as described in Agnor & Asphaug 2004). This, coupled with the fact that TNOs are significantly fainter than main belt asteroids, means that such detections will be significantly more difficult—and it is actually quite remarkable that the first TNO family should have been found so soon after the discovery of the first members of the population.

4.1.3. Giant Impacts and Planetary Satellites

Even within the satellite systems of the solar system, there is ample evidence for the key role played by giant impacts in shaping the evolution of the system. A number of the satellites bear the scars of impacts that were close to totally disrupting the body—the largest crater on Saturn's moon, Mimas, for example, Herschel, is around a third of the diameter of the moon, and the collision which formed it left fractures that can be found all around the satellite. Were the impactor slightly more massive, or traveling a little more quickly, it could well have totally disrupted the moon (e.g., Beatty et al. 1981).

Staying within the Saturnian system, giant impacts have been proposed as the source of the unusual ridge which encircles another of Saturn's moons, Iapetus. The ridge, which runs along the satellite's equator, is one of the most unusual looking features of the solar system, leaving the moon resembling a cricket ball with a raised seam. In this case, however, the seam is over 10 km high, and some 20 km wide, and stretches almost completely around the circumference of the satellite. Over the years, a number of explanations have been suggested for the ridge, ranging from it being a relic of the satellite having once been significantly oblate (e.g., Castillo-Rogez et al. 2007), to the collapse of a circum-Iapetan ring (Ip 2006; Stickle & Roberts 2018), or even the contraction of the moon's interior as it cooled following its formation (e.g., Sandwell & Schubert 2010). Dombard et al. (2010, 2012) have proposed that the feature could be the result of a giant impact upon the satellite. They suggest that a giant impact upon the satellite led to the formation of a moon, in orbit around Iapetus, in much the same way that models suggest our own Moon, and Pluto's Charon (among others) were formed. Such an object would accrete on an orbit close to the equatorial plane of the satellite, and then undergo tidal decay, with its orbit shrinking until it passed within the Roche limit of Iapetus, at which point the tides raised upon it by the large satellite would tear it into fragments, which would then slowly de-orbit to the surface of Iapetus at subsonic speeds. This would allow sufficient material to fall to create the ridge while simultaneously explaining the fact the ridge lies on the satellite's equator. More recently, Leleu et al. (2018) suggest that the merger of two similarly sized objects could explain not only Iapetus' unusual equatorial ridge and oblate shape, but that similar processes could well be behind the peculiar shapes of several of Saturn's other innermost satellites—Pan, Atlas and Prometheus.

A giant collision leading to the disruption of a massive, close-in Saturnian satellite has also been proposed as the origin of the planet's splendid ring system (e.g., Ip 1988). More recently, Charnoz et al. (2009) take this idea one step further, suggesting that Saturn's rings were formed during the putative Late Heavy Bombardment (which we discuss in more detail in Section 4.2.1). In that work, they suggest the bombardment would also have led to significant disruption of the closer in Saturnian satellites, with all smaller than Mimas being destroyed, and then re-accreting after the bombardment finished—which in turn suggests that, if that theory is to be believed, the giant Herschel crater could be the relic of an impact during that epoch that almost managed to disrupt Mimas. Instead of invoking a giant collision, Canup (2010) suggests that the rings could instead have been formed by the tidal stripping of an inward-spiraling Titan-sized satellite, during the final stages of the planet's formation. If such a satellite were to migrate inwards through a circum-planetary disk surrounding Saturn, Canup found that Saturn could strip the mantle and crust from such an object, while leaving its core relatively intact, fated to spiral inwards and be devoured by the planet. Given the size of such a satellite, it would be reasonable to expect it to be fully differentiated, and hence the material stripped away in this manner would be a good match to the observed composition of Saturn's rings—which are thought to be between 90% and 95% water ice. However, recent results based on analysis of data obtained during the final phase of the Cassini mission have cast doubt on both these proposed origins for the Saturnian ring system. Rather than being an ancient feature, dating back to the solar system's youth, that data instead suggests that the rings are far younger than previously thought—having formed in just the last 10–100 million years (Iess et al. 2019). Indeed, it seems likely that the rings are actually a transient feature—with observations suggesting that they may well decay within the next 300 million years (e.g., O'Donoghue et al. 2019). Such a recent origin is still entirely compatible with the idea that the rings were created in a catastrophic collision—and the transient nature of the Saturnian ring system seems in keeping with the fact that the other three giant planets, Jupiter, Uranus and Neptune, each also have ring systems—albeit ones far less spectacular than those that gird Saturn.

In orbit around Uranus, the small satellite Miranda proved the highlight of the Voyager 2 flypast for many astronomers. Unlike the other Uranian satellites imaged, which were heavily cratered, pictures of Miranda revealed an incredibly unusual terrain—as can be seen in Figure 13.

Zoom In Zoom Out Reset image size

Some of the features looked as though large "blocks" had been dropped into the planet, with adjacent regions appearing greatly distinct from one another. Some of the surface appeared "normal," with the usual craters, while other regions exhibited stretch marks, and enormous cracks stretched across other parts of the satellite. One particularly fascinating feature was the region known as Verona Rupes, a giant vertical cliff face that may be as much as twenty kilometers in height—the tallest cliff in the solar system. In order to explain these features, it has been suggested that Miranda was formed through the re-accretion of debris left over when an earlier moon in its orbit was collisionally shattered (e.g., McKinnon 1988; Marzari et al. 1998; Movshovitz et al. 2015). Although this theory remains the most favored in discussions of Miranda's formation, some authors (e.g., Marcialis & Greenberg 1987; Janes & Melosh 1988; Pappalardo et al. 1997; Hammond & Barr 2014) have suggested that other mechanisms (such as convection-driven resurfacing) could create the observed features, removing the need for their cataclysmic origin.

The families of irregular satellites in orbit around the outer planets are another example of the effects of giant collisions. In much the same way as the asteroid families mentioned in Section 4.1.2, distinct families of irregular satellites can be found around the giant planets, each of which is thought to be linked to the collisional disruption of a larger parent object, followed by gentle orbital dispersion (e.g., Sheppard & Jewitt 2003; Nesvorný et al. 2004), a finding supported by recent cladistical analysis of the satellite systems (Holt et al. 2018). Despite this collisional erosion of the irregular satellite population, it has been noted that the number of irregular satellites larger than a given size appears to be roughly constant between the outer planets (e.g., Jewitt & Sheppard 2005). Jewitt & Haghighipour (2007) present a detailed review of our understanding of these unusual objects, and describe the various models that have been proposed for their origin in some detail. Future observations of these satellites, and the discovery of further members of the population, will not only help to constrain the mechanism by which they were captured, but will also help to constrain the exact processes involved in the formation of their host planets themselves.

Neptune's gigantic irregular satellite Triton is an oddity among the planetary satellites. The seventh largest satellite in the solar system (with a diameter of 2706 km), it is by far the largest and most massive irregular satellite known, with a diameter an order of magnitude larger than Saturn's Phoebe, the next largest irregular. It also orbits remarkably close to its host planet, compared to the other irregulars, with an orbital radius of 3.5 × 10⁵ km placing it slightly closer to Neptune than our Moon is to the Earth (by contrast, the next closest irregular satellite known is Uranus' Francisco, with an orbital radius of just over 4 × 10⁶ km). Triton's unusual properties mark it as having a markedly different origin to the other irregular satellites, and many theories have been suggested over the years to explain its current orbit. It is noteworthy that, aside from Triton, Neptune lacks any sizable satellites—of the thirteen moons in the Neptunian system, Triton contains over 99% of the total mass. The second largest of Neptune's moons, Proteus, has dimensions of just 424 × 390 × 396 km (Stooke 1994), and the total mass of the regular satellite system of Neptune is far, far smaller than that of any of the other giant planets. In addition, the orbit of the irregular satellite exterior to Triton, Nereid, is remarkably eccentric.

All this evidence points to an unusual and possibly catastrophic origin for Triton (e.g., Banfield & Murray 1992). It is widely acknowledged that Triton is most likely a captured object, rather than one which formed in its current orbit, but the mechanism by which it was captured is still under some debate. Early theories (e.g., Lyttleton 1936b) attempted to link the origin of Triton to that of Pluto. This was based on the observational similarities of the two objects, the fact that Pluto's orbit crosses that of Neptune, and on estimates that both objects were of order 1/10th the mass of the Earth. After a lengthy discussion of hypothetical encounters between Neptune, Pluto and Triton, Lyttleton (1936b) highlighted a scenario by which such an encounter could capture Pluto to a prograde orbit around Neptune while simultaneously causing the direction of Triton's orbit to reverse, and also become prograde. Noting that dynamics is a time-reversible process, Lyttleton then proposed that Triton and Pluto could initially have been large prograde satellites of the giant planet. At some point, such an encounter, operating in reverse, could have occurred, leading to the ejection of Pluto to its current orbit, and could have been the origin of Triton's current retrograde orbit. The idea that Pluto was once a Neptunian satellite has played an important role in many theories of the origin of Triton—from the proposed disruption of the Neptunian satellite system by an encounter with a massive trans-Neptunian planet (Harrington & van Flandern 1979), to suggestions that Triton was an unbound interloper that was captured by Neptune through a close encounter with proto-Pluto (suggesting also that the approach was so close that the proto-Pluto was disrupted in the process, leading to the formation of its largest moon, Charon; Farinella et al. 1979). It should be noted, however, that none of these theories are currently considered as serious answers to the question of either Triton or Pluto's origin—but they still serve to highlight how our knowledge of Pluto and Triton has changed over the past century, and how those changes have informed our theories on their origins.

As our knowledge of both Triton and Pluto increased, it was realized that their masses were so low that such scenarios were implausible, unless the two objects had been involved in an actual physical collision (e.g., McKinnon 1984). Once this was realized, studies instead typically concentrated on mechanisms through which Triton could be captured alone—although some authors (e.g., Goldreich et al. 1989; Woolfson 1999, 2013) continue to consider the statistically improbable idea that Triton was captured as the result of such a giant collision. The model currently accepted by the majority of researchers, however, considers that Triton was likely a binary dwarf planet (like Pluto, or Eris) that approached Neptune sufficiently closely that it was captured by the planet as a result of the disruption of the binary pair—with the other member of the system continuing on an unbound orbit (e.g., Agnor & Hamilton 2006). Rufu & Canup (2017) considered the effect that such a capture event would have on a pre-existing satellite system, and found that the interactions between a newly captured Triton and a satellite system comparable to that seen orbiting Uranus could readily result in a system like that seen around the modern-day Neptune. In addition, they note that the interactions involved in such a scenario offer a mechanism by which the orbital eccentricity of Triton could be reduced on a timescale short enough to prevent the ejection of satellites like Nereid, which might otherwise be lost from the system as a result of scattering by Triton should its orbit solely be circularized by tidal interactions with Neptune.

At sizes smaller than those at which the impactor would destroy the target body, impacts become ever more frequent. While the epoch of giant collisions, at least from the point of view of the planets, passed in the early days of the solar system, smaller impacts have continued right up to the current day. Although such impacts are no doubt of interest when one considers the potential habitability of Earth-like planets (e.g., Horner & Jones 2008a, 2010a, 2011, 2012), the effects of such impacts, taken individually, would be so minor as to be undetectable in other planetary systems. Indeed, the first confirmed impacts to be observed on any solar system body only came in the last thirty years⁴³—the Shoemaker–Levy 9 impacts on Jupiter in 1994 (e.g., Hammel et al. 1995; Asphaug & Benz 1996), the two further impacts upon that planet observed in 2009 and 2010 (e.g., Sánchez-Lavega et al. 2010; Orton et al. 2011), and the impact of meteoroids observed on the surface of the Moon (e.g., Bellot Rubio et al. 2000; Ortiz et al. 2002; Madiedo et al. 2014, 2019).

Despite this, the study of these impacts in our solar system has provided a wealth of information on the processes by which the objects within formed and evolved. Studies of the chemistry and origin of meteorites reveal the timescales on which asteroids accreted and differentiated (e.g., Whitby et al. 2000; Kleine et al. 2002, 2005), as well as providing strong evidence that our solar system was seeded with short-lived radionuclides by a nearby supernova at some point during its formation (e.g., Lattimer et al. 1978; Hoppe et al. 1996; Yin et al. 2002; Mostefaoui et al. 2005). In addition to these results, which clearly provide valuable data to be incorporated in models of planetary formation, the two main areas of small-impact research that probably have the most impact on our studies of exoplanets are the proposed Late Heavy Bombardment of the terrestrial planets, and the suggested role of giant planets in shielding terrestrial planets from potentially catastrophic fluxes of small bodies.

4.2.1. The Late Heavy Bombardment

4.2.2. Do Giant Planets Offer Shielding to Terrestrial Worlds?

As described earlier, there are three main types of potentially hazardous objects in our own solar system, each of which finds its origin in one of the three main reservoirs of dynamically stable objects. The near-Earth asteroids originate within the Asteroid belt (e.g., Binzel et al. 1992; Bottke et al. 2000; Morbidelli et al. 2002a; Dunn et al. 2013), between the orbits of Mars and Jupiter, and are injected into the inner solar system, after being created in collisions between larger bodies, primarily by distant perturbations from the planets Jupiter and Saturn. The short-period, or Jupiter-family, comets are continually replenished from the Centaur population (e.g., Horner et al. 2003, 2004a, 2004b; Tiscareno & Malhotra 2003; Volk & Malhotra 2008; Grazier et al. 2018, 2019; Peixinho et al. 2020), which are themselves sourced from objects in the trans-Neptunian region (e.g., Holman & Wisdom 1993; Duncan & Levison 1997; Emel'yanenko et al. 2005; Volk & Malhotra 2008), though a significant contribution to that population may originate among the Neptunian and Jovian Trojan clouds (e.g., Marzari et al. 1995b; Horner & Evans 2006; Horner & Lykawka 2010a; Horner et al. 2012c). The third population of potential impactors, the Oort cloud comets, are sourced from the Öpik–Oort cloud (e.g., Öpik 1932; Oort 1950), and are typically thrown into the inner solar system by the gravitational influence of passing stars and the galactic tide (e.g., Hills 1981, Biermann et al. 1983; Heisler et al. 1987; Matese & Lissauer 2004; Feng & Bailer-Jones 2015; Vokrouhlický et al. 2019), although it has been suggested that some fraction might well be sent inwards as a result of perturbations by a Jovian-mass Solar companion (e.g., Matese et al. 1999; Horner & Evans 2002; Matese & Whitmire 2011), as we will discuss in more detail in Section 4.5.1.

In the case of the near-Earth asteroids and short-period comets, despite the presence of relatively populous reservoirs of parent objects (the asteroid belt, and trans-Neptunian disk, respectively), the number of objects being moved from orbits in those regions to orbits in the inner solar system (where they can imperil the Earth) would be minimal were it not for the perturbative influence of the giant planets. For this reason, it is immediately apparent that the presence of giant planets can play a significant role in determining the frequency with which debris rains down upon the terrestrial planets. In other words—the presence of a massive debris disk around a given star should not be taken, per se, as evidence that the frequency of collisions on a planet in a different part of that planetary system would be prohibitively high for life to develop. If no giant planets were present in that system, then no mechanism would exist to effectively transfer material from that disk to a potentially habitable planet elsewhere in the system, and so one might instead expect the impact regime on such a world to be relatively peaceful, despite the presence of a massive disk.

Despite the clarity of such logic, it has historically been widely taught that the planet Jupiter acts as a giant shield, significantly lowering the rate at which asteroids and comets collide with the Earth. Without Jupiter, the argument goes, significant impacts would be sufficiently frequent on the Earth that the development of life would be stymied, or even prevented absolutely (e.g., Ward & Brownlee 2000). The origin of this idea is unknown, but it likely has its roots around the time that craters on the surface of the Earth were finally acknowledged as being the result of impacts—an idea that was not fully accepted until the 1960s (e.g., Chao et al. 1960; Shoemaker 1963; Shoemaker & Chao 1961; Bjork 1961).

Given that weathering removes all but the largest craters on relatively short geological timescales, it was obvious that most craters could not be ancient features, but had instead been created by impacts in the relatively recent past. At that time, very few short-period comets, and even fewer near-Earth asteroids were known. The obvious culprits, therefore, were the long-period comets, objects sourced from the Oort cloud. Such comets are only tenuously bound to the Sun, with their aphelia located at enormous distances (typically tens of thousands of au), and as such only need to experience a very small perturbation in order to become unbound, and be ejected from the solar system forever. In most cases, the planet that ejects comets in this way is Jupiter, and so therefore it seems an obvious conclusion that, if the planet were not there, far more of these objects would survive to pass through the inner solar system multiple times, and therefore the impact rate at the Earth would be much higher. The idea that Jupiter has shielded us is now firmly entrenched in both the public and scientific consciousness. By extension, it is often argued that giant planets are a necessity in order to shield Earth-like planets from impacts for those planets to be habitable (as in, for example, Greaves 2006, who suggests that the lack of a Jupiter-type planet in the Tau Ceti system has resulted in a significant amount of debris being present that would threaten any Earth-like planets therein).

Given the pervasiveness of this idea, it is somewhat surprising that, until recently, very little work had been done to examine it in any depth. The first study of any note into the question of planetary shielding was carried out by Wetherill (1994), who concluded that, in planetary systems containing only "failed Jupiters" (i.e., planets which only manage to reach the mass of Uranus or Neptune, rather than that of a Jupiter), the impact flux experienced by an Earth-like planet would be of order a thousand times higher than that experienced in our own solar system. This would be a direct result of the reduced efficiency with which small objects would be ejected from such a system, allowing them to remain on planet-crossing orbits for a much longer period of time. However, in order for a given object to pose a threat to an Earth-like planet, it must first be emplaced on an orbit that crosses that of the planet. The dynamical lifetime of such objects is relatively short, and so in order for small bodies to pose an on-going threat, their numbers must continually be replenished—a question which was not addressed in that early work.

Laakso et al. (2006) approached the question of planetary shielding from a different angle. Using numerical integration, they examined the effect of the position and mass of a Jovian planet on the rate of ejection of particles that were initially placed on eccentric orbits that crossed the classical habitable zone from the start of the integration. Using our solar system as a test case for their method, they found the surprising result that Jupiter "in its current orbit, may provide a minimal amount of protection to the Earth." Unfortunately, again, their work did not consider the transfer of material from non-threatening to dangerous orbits, but it did provide the first hints that something might be a little amiss with the canonical view of Jupiter's role.

Our current understanding of the various populations of bodies that can threaten the Earth is vastly different to that which was present around the time that the idea the Jupiter is a shield caught on. Rather than the long-period comets being the main source of impact threat to the Earth, it is now believed that the Near-Earth Asteroids constitute the great bulk of the impact risk, with the comets contributing just a few percent of the total risk (e.g., Chapman & Morrison 1994; Bottke et al. 2002). While it is still true that Jupiter plays a significant role in removing the near-Earth asteroids from threatening orbits, it is also well known that it is the main source of fresh asteroidal material in the inner solar system (e.g., Morbidelli et al. 2002a, 2002b). In the case of the short-period comets (often known as the Jupiter-family), the situation is equally confused—the great majority of such comets make their way onto Earth-crossing orbits as a result of Jupiter's perturbations, and so any conclusions as to the true level of shielding offered by the giant planet must take into account the efficiency with which it creates new threats, as well as the ease with which it removes them.

In order to study this problem, a series of detailed dynamical studies have been carried out (Horner & Jones 2008a, 2008b, 2009, 2010b, 2012; Horner et al. 2010) that examine the influence of Jupiter on the three populations of potentially hazardous objects in our own solar system—namely the near-Earth asteroids, the short-period comets and the long-period comets. The results are startling. In the case of the near-Earth asteroids, it turns out that the Earth actually experiences a greater number of impacts with our Jupiter than it would with no Jupiter at all (see Figure 14, below). However, if Jupiter were just a quarter of its current mass, the impact rate of near-Earth asteroids would be far higher than in the other two cases. This rise and fall of the impact rate as the mass of Jupiter increases is found to be a direct result of the changing location of the ν₆ secular resonance, which can drive asteroids trapped within into the inner solar system on astronomically short timescales. In our own solar system, this resonance, although strong, is located almost at the inner edge of the asteroid belt, where very few objects can fall prey to its perturbative influence. When Jupiter is less massive, however, the resonance is located farther from the Sun, and is also somewhat broader, and can therefore influence a greater fraction of the total mass of the asteroid belt. The "worst case" scenario occurs when Jupiter is around a quarter of its current mass (similar in mass to the planet Saturn), at which point the resonance is both broad and strong, and is located in the middle of the asteroid belt, ideally located to disturb the orbits of the maximum number of asteroids.⁴⁴

Zoom In Zoom Out Reset image size

For the short-period comets, a similar pattern emerges, albeit for different reasons—as can be seen in Figure 14. At low Jupiter masses, the planet is unable to either efficiently eject comets from the solar system or emplace them on Earth-crossing orbits. At high masses, the planet can very efficiently emplace comets on Earth-crossing orbits, giving them the opportunity to collide with our planet. However, it can also easily remove them from the inner solar system and eject them, so any given comet will most likely have relatively few opportunities to hit the Earth before being removed again. In both these cases, then, the impact flux experienced by Earth is relatively low. Once again, there is a "worst case" scenario, however. Remarkably, this "worst case" again occurs when the Jupiter-like planet has around a quarter of the mass of our Jupiter. At that mass, the planet can efficiently emplace comets on to Earth-crossing orbits. However, it is still of sufficiently low mass that it is very difficult for it to eject those comets from the solar system entirely. As such, any comets placed on Earth-crossing orbits by the giant planet at that mass will have plenty of opportunities to collide with the Earth before they are removed, and so the impact flux reaches a peak. Beyond that peak, as the planet's mass grows larger, objects are ever more efficiently ejected, causing the time spent on Earth-crossing orbits to fall, thereby reducing the impact rate.

In the case of the long-period comets, it seems that the long-held belief holds true—Jupiter does indeed act as a shield. As its mass increases, it becomes ever more efficient at perturbing loosely bound comets in such a way that their orbits become unbound, and they head out of the system, never to return. A more detailed study of the effect of giant planets on the long-period comet flux (Lewis et al. 2013) studied the evolution of cometary orbits from the initial population of the Oort Cloud by the giant planets through to their eventual re-injection and ejection from the inner planetary region. That work once again highlighted the role of giant planets (of mass comparable to, or greater than, Saturn) in ejecting potentially Earth-crossing comets from the system, ensuring that the chance of them colliding with a terrestrial planet is minimized. Once again, for long-period comets, it seems that Jupiter (and Saturn) truly act to reduce the impact threat posed to the Earth.

Although a significant amount of further research into the question of the shielding offered by giant planets is needed before any definitive answer is reached as to the true balance between their roles as friend and foe, it is clear from these studies that, at the very least, the situation is significantly more complicated than had previously been thought (e.g., Grazier 2016; Grazier et al. 2018). Rather than being the benign and beneficial influence on our planet's development that has long been assumed, it seems that Jupiter might actually be more of an enemy than a friend!

The idea that giant planets migrate during the process of their formation and evolution is now well accepted. In exoplanetary systems, key evidence for this behavior includes the hot Jupiters (e.g., Mayor & Queloz 1995; Johnson et al. 2006; Vogt et al. 2010; Winn et al. 2010; Bryan et al. 2012; Becker et al. 2015), and a number of dynamically compact systems with planets trapped on mutually resonant orbits (e.g., Johnson et al. 2011; Lissauer et al. 2011; Robertson et al. 2012a, 2012b; Wittenmyer et al. 2012b; Gillon et al. 2017). Such migration is the most common explanation for all giant planets discovered orbiting closer to their host stars than the location of the ice line, with the idea being that the planet first accreted beyond that point in its host system before careening inwards during the latter stages of its formation (e.g., Kuchner & Lecar 2002; Masset & Paploizou 2003; Crida & Morbidelli 2007; Beaugé & Nesvorný 2012)—although in recent years, alternative mechanisms have been proposed to explain at least some of the known short-period giant planets (e.g., Batygin et al. 2016; Bailey & Batygin 2018; Dawson & Johnson 2018). It has even been suggested that such inward migration might often lead to the in-fall of one (or more) giant planets to the atmosphere of a youthful star, polluting that star's atmosphere with the heavy elements accreted, and so artificially inflating the observed metallicity (e.g., Pasquini et al. 2007). A planet migrating in such a fashion might also drag with it vast quantities of volatile material from beyond the ice-line, leading to the accretion of "water-worlds," which might potentially be habitable, in the giant planet's wake (e.g., Fogg & Nelson 2007a, 2007b, 2009; Izidoro et al. 2014b; Darriba et al. 2017).

Within our own solar system, too, evidence abounds for the post-formation migration of the giant planets. However, while currently migration scenarios involving exoplanets typically only discuss inward migration, within our own solar system there is a significant body of evidence that argues for the outward migration of Neptune, and potentially Uranus, leading to the dynamical relaxation of the outer solar system. Indeed, it is currently thought that Neptune migrated over a distance of at least 7 au, and potentially as much as 15 au outward after its formation, doubling the radial extent of the planetary component of our solar system.

4.3.1. Evidence for Migration—The Small Bodies

4.3.2. Migration and the Late Heavy Bombardment

One of the ingredients considered most vital for the development and ongoing survival of life is water (e.g., McKay 1991, 2014; Rampino & Caldeira 1994; Chyba & McDonald 1995; Ward & Brownlee 2000; and many others). In fact, the presence of water is considered so important that the classical definition of the "Habitable Zone" around a star is simply the range of distances from the star over which an Earth-like planet would be able to have stable liquid water at some point on its surface (e.g., Hart 1979; Kasting et al. 1993; Horner & Jones 2010b; Kopparapu et al. 2013, 2014, 2016; Zsom et al. 2013; Agnew et al. 2017).

The role of water is considered to go beyond simply providing a medium in which life can develop, and being the key solvent for its survival. It is thought, for example, that if the Earth was desiccated, plate tectonics would not occur, which would remove a key route through which gases that are removed from the atmosphere by weathering, life, and other processes, are recycled from the surface (e.g., Wallace & Hobbs 2006, Section 2.3). It is even argued that the lack of plate tectonics might lead to the collapse of the dynamo that produces the magnetic field that protects our planet (as is the case for both Venus and Mars; e.g., Nimmo & McKenzie 1998; Nimmo 2002; Breuer & Spohn 2003; O'Neill et al. 2007). For a more detailed review of the various influences on planetary habitability, we direct the interested reader to the detailed review of that topic, Horner & Jones (2010b).

Given that water is such an important part of life on Earth, it is somewhat surprising that the origin of that water is still under heavy debate. Our planet formed well sunwards of the location of the ice-line in the solar system, and so temperatures were far too high for water to condense from the Solar nebula to be directly accreted in situ. It seems obvious, then, to conclude that the terrestrial planets should have accreted solely from dry rocks. It has been proposed, however, that a significant amount of water could be trapped in the form of hydrated silicates in that region, which would allow water to be accreted concomitantly with the planet's silicate budget. The apparent paradox of our wet planet, and its hot origins, has prompted significant amounts of debate, and has led to a number of different models being proposed to detail the hydration of the terrestrial planets. These can be broken down into three main types—endogenous hydration, early exogenous hydration and "late veneer" exogenous hydration.

Models detailing the endogenous hydration of the terrestrial planets suggest that the great bulk of their water was accreted from their local environment, with only a small contribution being added from sources farther from the Sun. In such a scenario, the planets are thought to have accreted the bulk of their water in the form of hydrated silicates, which have been shown to be stable at far higher temperatures than those at which water can exist alone as a solid (e.g., Drake 2004, 2005; Asaduzzaman et al. 2015). One problem with such scenarios, which on the surface seem to provide a promising avenue by which terrestrial planets throughout the universe could gain significant amounts of water during their formation, is that there is a well established correlation between the water content of meteorites that fall on the Earth and the region of the asteroid belt in which they originated (with those that originated farther out in the belt having a greater water content than those with origins in its inner reaches). In particular, Morbidelli et al. (2000) detail results for a class of meteorites known as the Enstatite chondrites, which are thought to originate in the innermost region of the asteroid belt. They highlight the fact that these meteorites are the driest of all known meteorites in the solar system (containing only 0.05%–0.1% water, by mass), which suggests that the rocky material which formed interior to the ice-line was unable to hold much in the way of water, and therefore that the amount of water concomitantly accreted during the formation of the terrestrial planets from local materials most likely was insufficient to explain the Earth's water budget.

Because of these problems tying the Earth's water to the materials from which it formed, most researchers now believe that exogenous models of hydration are the best explanation for our water budget. Such scenarios postulate that, rather than having a local reservoir of hydrated material, the water budget of the terrestrial planets was instead sourced from objects that formed much farther from the Sun, beyond the ice-line, and were then flung inwards to impact on the planets. These models are further broken down into two main types—early accretion models, which suggest that the water was pumped into the inner solar system (typically from objects that formed in the outer reaches of the asteroid belt, just beyond the ice line) while the planets were themselves in the process of accreting (e.g., Morbidelli et al. 2000; Petit et al. 2001; Marty 2012). Some of these models require the bulk of the water to have been delivered in a few stochastic giant impacts (similar to those described in Section 4.1; e.g., Marty 2012), a process which has been invoked to explain the fact that the terrestrial planets appear to have significantly different water budgets (with, for example, the Earth being significantly more hydrated than Mars). It has even recently been suggested that the bulk of Earth's water was delivered in the planetary collision that gave birth to the Moon—with Earth's water being delivered by the impacting planet, Theia (Budde et al. 2019). On the other hand, mechanisms have also been proposed that could deliver water in the form of a continuous early rain of icy pebbles, ranging in size from millimeters to tens of centimeters in diameter, that formed through pebble accretion beyond the ice-line, then migrated inward following the ice-line as it, in turn, briefly moved in to heliocentric distances smaller than that of the Earth's orbit (e.g., Sato et al. 2016).

The "late-veneer" exogenous hydration models, by contrast, suggest that the hydration of the terrestrial planets occurred at some point after their bulk accretion was completed (e.g., Chyba 1987; Owen & Bar-Nun 1995). One potential event which could have been the source of this hydration is the Late Heavy Bombardment (e.g., Oberbeck & Fogleman 1989; Wells et al. 2003), which may well have led to a prolonged influx of material from both the outer Asteroid belt, and the trans-Neptunian disk, through the inner solar system (e.g., Gomes et al. 2005). Such a bombardment, even though it would consist of many impacts of smaller bodies, rather than a few giant impacts (as in the early exogenous scenarios) could easily lead to the terrestrial planets having different water budgets, particularly if the impactors were sourced from both the Asteroid belt and the outer reaches of the solar system, as discussed by Horner et al. (2009). Equally, the Grand Tack model, in which Jupiter and Saturn careen inward through the asteroid belt, and then migrate back out to their current locations, could efficiently deliver material from beyond the snow-line to the terrestrial planets (e.g., O'Brien et al. 2018).

One way in which researchers have attempted to differentiate between the various methods suggested for the delivery of water to the Earth is through the study of the ratio between deuterium and hydrogen in water (the D:H ratio). A number of studies (e.g., Drouart et al. 1999; Dauphas et al. 2000; Mousis et al. 2000; Mousis 2004; Horner et al. 2007, 2008; Yang et al. 2013) of the formation of our solar system have suggested that the D:H ratio in the water accreted to icy bodies should vary significantly as a function of heliocentric distance within the Solar nebula—with the lowest values being found near the ice-line (in the outer Asteroid belt), and the largest values occurring in the region beyond Neptune. Clearly, then, one should expect that water delivered from the asteroid belt would have a measurably different D:H signature than that sourced from the outer regions of the solar system—a result supported by observed large D:H values in comets (e.g., Balsiger et al. 1995; Eberhardt et al. 1995; Bockelée-Morvan et al. 1998; Meier et al. 1998; Villanueva et al. 2009), and the significantly lower values measured in meteorites (e.g., Figure 2 of Drouart et al. 1999). One would expect that it would be a relatively simple procedure to compare the different values expected for different objects to that measured in the Earth's oceans, and therefore have an independent test of the origin of the water stored therein. However, the issue is complicated by the fact that the D:H ratio measured in the Earth's water is not solely the product of its origin, but rather has been significantly modified over geological timescales by a variety of processes. This also greatly hinders comparisons between the deuteration of terrestrial, venutian, and Martian water. For simplicity, many authors assume that the initial D:H ratio of water on these planets was identical (e.g., Gurwell 1995; Krasnopolsky et al. 1998; Lécuyer et al. 2000) and then use the observed differences in the deuteration at the current day to draw conclusions on the processes that have occurred on those planets in the intervening time.

Such difficulties are one of the key reasons that the true origin of the Earth's water has not yet been universally agreed, although, at the current time, the most widely accepted scenario seems to be that the bulk of the water was sourced from far beyond the Earth's orbit, with at least some contribution coming as a "late veneer" during the Late-Heavy Bombardment (e.g., Morbidelli et al. 2012).

The situation is further complicated by the fact that Venus is markedly drier than the Earth at the current epoch, while Mars is believed to have once been both warm, and wet (although most of that water is likely still locked up in permafrost across the planet). It seems likely that both those planets have lost significant amounts of water through it being stripped from their atmospheres (e.g., Jakosky & Phillips 2001; Hodges 2002; Lammer et al. 2003; Valeille et al. 2010). In addition, a number of studies have suggested that both Mars and Venus were dessicated as a result of the solar system's early impact regime (e.g., Davies 2008; Kurosawa 2015; Rickman et al. 2019b). While the final narrative of the inner solar system's hydration remains to be determined, it seems increasingly likely that the giant collisions that dominated the latter stages of terrestrial planet formation may have played a vital role in ensuring the disparate volatile budgets of Venus, Earth and Mars (e.g., Lykawka & Ito 2019).

Ever since Neptune was discovered as a direct result of its perturbations on the orbit of Uranus, the idea that there might be one (or more) planetary body orbiting farther out has repeatedly come into vogue. The fortuitous discovery of Pluto, in 1930, was a direct result of a search for such a body—a search that simply led to the detection of the first trans-Neptunian object. In the early 1980s, studies of the frequency of mass extinctions on the Earth, allied to the discovery of the Chicxulub impact crater in Mexico, believed to be tied to the extinction of the dinosaurs (e.g., Alvarez et al. 1980; Hildebrand et al. 1991; Schulte et al. 2010), led to the idea that such extinctions were occurring with a period of approximately 26 million years (e.g., Raup & Sepkoski 1984, 1986; Schwartz & James 1984). In order to explain the observed periodicity, a number of competing theories sprang up, including the idea that our Sun might have a distant companion, most likely a brown or red dwarf, moving on a highly eccentric orbit that would, at perihelion, bring it through a sufficiently dense region of the inner Oort cloud that it would trigger a severe shower of comets, leading to impacts on the Earth and a mass extinction (e.g., Alvarez & Muller 1984; Davis et al. 1984; Whitmire & Jackson 1984). In the years that followed, however, the theory of "Nemesis" quickly fell by the wayside, with a number of studies casting doubt on the validity of the hypothesis (e.g., Bailey 1984; Clube & Napier 1984a; Hills 1984; Hut 1984; Torbett & Smoluchowski 1984). However, theories of planetary-mass objects beyond the orbit of Neptune continue to crop up in order to explain a variety of features of our solar system, from the population of detached objects beyond Neptune's orbit to observed aphelion asymmetries among dynamically new long period comets. Here, we detail a variety of those theories, showing how observations of the dynamical structure of the outer solar system can allow us to make predictions about what may lie at greater distances.

4.5.1. A Planet in the Oort Cloud—Source of the Long-period Comet Distribution?

4.5.2. A Trans-Neptunian Planet as an Explanation for the Structure of the Trans-Neptunian Population

Stretching halfway to the nearest star, the Oort cloud is one of the great mysteries of the solar system. While we know it is there, as a result of the continual flux of dynamically new comets maintaining the long-period comet population (e.g., Öpik 1932; Oort 1950), it is too distant to study directly. Instead, we must learn what we can from the long period comets it sends our way.

As we discussed earlier, those comets are injected from the Oort cloud through the combined effects of the galactic tide, and perturbations from passing stars (e.g., Feng & Bailer-Jones 2015, and references therein), In coming years, it is likely that the combined efforts of the Gaia space telescope (Gaia Collaboration et al. 2016, 2018; Evans et al. 2018), surveying the motions and distances of approximately 2% of all stars in the galaxy, and the next generation of all sky surveys (such as the Large Scale Synoptic Telescope, LSST; LSST Science Collaboration et al. 2009; Ivezić al., 2019), which will discover vast numbers of new comets, will reveal evidence of additional past encounters between the Sun and passing stars, to add to the Biermann comet shower (Biermann et al. 1983). Within a decade, we may even get our first close-up view of a dynamically new comet, passing through the inner solar system for the first time in more than four billion years, as a result of the recently announced Eurpean Space Agency mission "Comet Interceptor,"⁴⁹ scheduled for launch in 2028.

But what of the Oort Cloud's origins? It is clear that it can not have formed in situ—instead, it is thought to have been created as a side-effect of the accretion of the giant planets, and their subsequent clearing of the outer solar system. As the giant planets accreted, they were embedded in a disk containing icy planetesimals of all shapes and sizes. Close encounters between the planets and the planetesimals would scatter the latter, and as the mass of the planets grew, so too did the efficiency with which such scattering events could eject those planetesimals to highly eccentric, and even unbound, orbits.

The farther an object strays from the Sun, the weaker the Sun's gravitational influence upon it will be. Once a comet is sufficiently distant, the influence of nearby stars and the galactic tide will therefore be strong enough to significantly alter the orbit of that comet. As a result, there are three possible outcomes for a planetesimal ejected by the giant planets to a high eccentricity orbit, that takes them well beyond the orbit of Neptune. The first is that they are either ejected with such great velocity that they move on a hyperbolic orbit, never to return, or are flung sufficiently far (on a nominally bound orbit) that the influence of nearby stars and the galactic tide can decouple them from the Sun's influence. Such objects, clearly, will not become members of the Oort Cloud. At the other extreme, you have those comets whose aphelia are not sufficiently far from the Sun for passing stars and the galactic tide to significantly perturb them. Those objects will return to the regime of the giant planets unperturbed, no doubt to experience further scattering. Between the two extremes, you have a scenario whereby an ejected object obtains sufficient distance from the Sun for its orbit to be strongly perturbed by the influence of the galactic tide and passing stars, without becoming so distant that those perturbations remove it from the solar system entirely. It is these objects that give birth to the Oort Cloud. The perturbations from the galactic tide and nearby stars act to lift the perihelia of those objects out of the region of the solar system dominated by the planets, placing them on orbits where they can remain for billions of years, held in cold storage.

Over the billions of years that follow, the orbits of comets in the outer reaches of the Oort Cloud will slowly become randomized, as a result of their continual stirring by the galactic tide, and periodic nudges from passing stars. From a distribution that would initially resemble the disk from which those comets were sourced (i.e., relatively low inclinations with respect to the plane of the solar system), the Oort Cloud will eventually obtain its current roughly spheroidal shape, with the inclinations of the comets therein wholly randomized.

The outer boundary, beyond which objects would have been effectively removed from the Sun's influence, would have been closer to the Sun if our birth cluster was particularly massive and dense (with more stars forming closer to the solar system), and at a greater distance if the solar system formed in a relatively loose cluster or stellar association. As noted at the start of Section 4, evidence gathered from meteorites suggests that our birth cluster was relatively massive, containing at least 2000 stars (e.g., Pfalzner et al. 2015). Such a tight cluster is invoked in order to explain the abundance of ²⁶Al in the early solar system, which it is argued was provided by the wind of a nearby massive star.

Dones et al. (2004) provide an excellent review of theories describing the formation and evolution of the Oort Cloud. They note that, on the basis of dynamical simulations, Uranus and Neptune likely dominated the emplacement of cometary material to the Oort Cloud—see their section four for a full overview of studies of the Oort Cloud's formation prior to that date. However, more recent work has reopened the question of the primary source of material to the primordial Oort Cloud. Brasser et al. (2012) performed detailed numerical simulations of the formation of the Oort, taking account of the influence of the other stars in the Sun's birth cluster. They found that the inner Oort cloud was primarily populated by comets ejected by Jupiter and Saturn, with the pericentres of those comets then being raised by encounters with passing stars to decouple them from the planetary domain of the solar system. As such, the question of the "planetary parents" of the Oort cloud remains somewhat in doubt, and the true answer might well depend on the precise nature of the cluster in which the Sun formed, as well as the chaotic migration history of the giant planets. For a more detailed discussion of the Oort cloud's formation, we direct the interested reader to the recent review by Dones et al. (2015), which builds on their earlier work to provide a detailed overview of the formation of all of the solar system's reservoirs of cometary body.

As discussed in Section 4.4, it has been proposed that the level of deuteration trapped in the water of the solar system's icy bodies could act as a tracer of their formation locations within that disk (e.g., Drouart et al. 1999; Dauphas et al. 2000; Mousis et al. 2000; Mousis 2004; Horner et al. 2007, 2008; Yang et al. 2013). As such, Horner et al. (2007) suggested that measurements of the D:H ratio in incoming comets could be used to test the various theories of their formation, showing how a variety of scenarios for the formation regions of cometary bodies would manifest as different distributions of deuteration in the nuclei we observe today. Following that logic, Kavelaars et al. (2011) compared the D:H enrichment measured in Saturn's icy satellite Enceladus with the levels of deuteration measured for a number of long-period comets. They found that the comets displayed similar levels of deuteration to the icy moon, suggesting that they had formed at roughly equivalent locations in the protoplanetary nebula. Given that the established wisdom is that the Oort Cloud was probably primarily populated by Uranus and Neptune, those authors considered this to be evidence that the majority of material in the Oort Cloud formed at a similar heliocentric distance to Saturn, which in turn would suggest that Saturn formed close to Uranus and Neptune. Such a scenario is exactly what was proposed by Thommes et al. (1999, 2002), and that formed the cornerstone of the Nice Model (Gomes et al. 2005; Morbidelli et al. 2005; Tsiganis et al. 2005).

Because of the tenuous grasp with which the Sun holds on to comets in the Oort cloud (particularly the outer reaches), the cloud should continually be bleeding objects to interstellar space, stripped from the Sun by the combined effects of passing stars and the galactic tide. It is fair to assume that other stars behave likewise—and a result, interstellar space should contain vast numbers of free-floating comets and asteroids, some of which, by chance, must occasionally pass through the inner solar system. The first such object, 1I/'Oumuamua, was discovered in 2017 (e.g., Bannister et al. 2017; Jewitt et al. 2017; Meech et al. 2017; Fitzsimmons et al. 2018)—and should another such object be discovered moving on a favorable path while the Comet Interceptor mission stands ready at Earth's L2 Lagrange point, that object would likely be considered an ideal target for the mission. In other words, in the relatively near future, that mission means that there is a chance that we may get our first "up-close-and-personal" view of a comet or asteroid from another planetary system.

Just as passing stars perturb our Oort cloud, so too will the Sun perturb the Oort clouds of its stellar neighbors. Particularly while the Sun was still young, and moving within its birth cluster, this process of mutual Oort cloud stirring would have been pronounced, and one could readily envision a cluster awash with comets, torn from one parent star, co-moving within the cluster. For this reason, Levison et al. (2010) performed numerical simulations to investigate the degree to which comets could be transferred from one star to another—with the idea that our Oort cloud could contain a significant number of comets captured from other stars before the dispersal of our birth cluster. The authors conclude that a significant fraction of comets in the Oort cloud were indeed captured from our Sun's birth neighbors—with such comets potentially making up the vast majority (>90%) of nuclei within the cloud.

While the capture of comets to the cloud might have been efficient while the Sun resided in its birth cluster, recent work by Hanse et al. (2018) has shown that the situation became markedly different once the Sun began wandering the galaxy alone, having left that cluster behind. They consider the capture of comets from field stars (and, by extension, the transfer of Solar comets to those stars) as those stars fly past the Sun. They find that capture events are relatively rare, since the typical encounter velocities of passing stars are far greater (by several orders of magnitude) than the orbital velocities of comets in the Oort cloud. They note that the transfer of any "exocomets" to the Oort cloud will typically only happen during unusually slow (≲0.5 km s⁻¹) and close (≲10⁵ au) encounters—and as a result, such exocomets likely make up just a small fraction of the mass of the cloud (between 10⁻⁴ and 10⁻⁵). In the same work, those authors were able to model how the Oort cloud is stripped of comets by stellar encounters, suggesting that such encounters have likely caused the Oort cloud to lose between 25% and 65% of its initial mass through the age of the solar system.

As we discussed in Section 4.5, comets from the Oort cloud have been invoked as the cause of periodic mass extinctions on Earth since the early 1980s. In order to explain the suspected ∼26 Myr periodicity in mass extinctions reported by Raup & Sepkoski (1984, 1986), several authors discussed the possibility that the Sun might have an undetected binary companion, a dim red dwarf star called "Nemesis," whose motion through the Oort cloud sent periodic showers of comets to imperil life on Earth (e.g., Alvarez & Muller 1984; Davis et al. 1984; Whitmire & Jackson 1984)—an idea that quickly fell into disfavor (e.g., Bailey 1984; Clube & Napier 1984a; Hills 1984; Hut 1984; Torbett & Smoluchowski 1984).

Schwartz & James (1984) proposed a different mechanism by which the Oort cloud could repeatedly send comet swarms toward the Earth, noting that the ∼26 Myr periodicity in extinctions found by Raup & Sepkoski (1984) was remarkably similar to the period with which the solar system oscillates about the plane of the galaxy. While it is typical to imagine the orbital motion of the Sun around the galaxy as being roughly circular (with an orbital period of ∼225 Myr), the true story is somewhat more complicated. As it moves around the galactic center, the Sun also undergoes a periodic vertical oscillation back and forth through the plane of the galaxy. When the Sun is located above that plane, the mass of stars, gas and dust within it exerts a downward force, pulling our star (and therefore the solar system) down toward the plane. As a result, the system accelerates, falling downwards until it passes through the galactic plane and out the other side. Once it is below the plane of the galaxy, the system now feels the same restoring force pulling upwards, causing its vertical motion to slow, stop, and then reverse. The result is that the Sun's motion around the galactic center is somewhat reminiscent of an automated fairground carousel ride, with the Sun oscillating up and down as it moves around the galactic center.

Schwartz & James (1984) proposed that it was this motion that was to blame for the periodic extinctions noted by Raup & Sepkoski (1984). The vertical oscillations of the Sun result in the solar system passing through the plane of the galaxy repeatedly, with those passages occurring with a frequency comparable to the periodicity proposed for Earth's mass extinctions. As the system passes through the galactic midplane, encounters with stars, and with giant molecular clouds become more likely, increasing in turn the likelihood that the Earth will collide with a comet flung inwards by such an encounter. The idea was revisited by Randall & Reece (2014), who invoked the influence of a smooth disk of dark matter, located at the midplane of the galactic disk, in order to explain how the Oort cloud could always be guaranteed to trigger a comet shower—something that had proved problematic when the only influences on the cloud were considered to be encounters with stars and giant molecular clouds.

An alternative theory connecting the Sun's motion through the galaxy was proposed by Leitch & Vasisht (1998), who suggested that several of Earth's largest extinction events coincided with periods at which the Sun was moving through one or other of our Galaxy's spiral arms. The spiral arms of the Milky Way (and of other spiral galaxies) are regions within which there is a greater concentration of dust, gas, and star formation—and as a result, a greater stellar density. When the Sun moves through one of those arms, then, the likelihood of close encounters with stars (which could inject comets to the inner solar system) is higher. At the same time, the chance of an extinction event caused by a nearby supernova (the death of one of the galaxies most massive stars) is also greatly enhanced, since those stars are almost exclusively found within the spiral arms. Gilman & Erenler (2008) and Filipovic et al. (2013) revisited this idea, using updated maps of the Milky Way's structure to attempt to determine whether any correlation between such spiral-arm crossings and mass extinctions could be found.

It is fair to say, however, that the jury is still out on whether mass extinctions on Earth are truly periodic, and even if they are, many of the theories proposed to explain such periodic extinctions are considered by many to be fanciful or unfeasible. Nevertheless, it is clear that comets from the Oort cloud have posed, and will continue to pose, an impact threat to the Earth—and the origin and evolution of the Oort cloud will remain a hot topic for research for many years to come.

When we look out at the solar system, we see dust everywhere. On any clear night, we observe dust ablating in Earth's atmosphere, producing meteors, and their brighter siblings fireballs and bolides (e.g., Jenniskens 2006). With the unaided eye, we can also observe the Zodiacal Light and Gegenschein (e.g., Roosen 1970; Leinert 1975), both the result of sunlight scattering from dust in the plane of the solar system. Whenever an active cometary nucleus is sufficiently close to the Sun, the sublimation of its surface ices will result in the continual ejection of dust grains—which will be pushed away from the Sun to produce those comet's spectacular dust tails. And when those comets fragment (as they often do; e.g., Crovisier et al. 1996; Sekanina 2000; Jenniskens & Lyytinen 2005; Jenniskens & Vaubaillon 2007), or go into outburst (e.g., Sekanina et al. 1992; Dello Russo et al. 2008; Montalto et al. 2008; Ye & Clark 2019), large quantities of dust are released—occasionally even resulting in spectacular meteor showers, as that dust encounters Earth (e.g., Asher 1999; Wiegert et al. 2005; Jenniskens & Vaubaillon 2007; Jenniskens 2006).

Asteroids, too, are a ready source of dust. The near-Earth asteroid 3200 Phaethon (the parent of the Geminid meteor shower; e.g., Gustafson 1989; Williams & Wu 1993) behaves as a "rock comet," shedding dust near perihelion as a result of surface weathering, caused by the extreme temperatures to which it is subject through the course of its orbit (e.g., Jewitt & Li 2010; Li & Jewitt 2013). More often, asteroids shed dust as a result of collisional processes—from the small scale sputtering caused by the continual hypervelocity impacts of dust on their surface, through to the rare but catastrophic collisions that shatter whole objects, creating vast quantities of dust in the process (e.g., Sykes & Greenberg 1986; Sykes 1990; Reach 1992; Durda & Dermott 1997; Nesvorný et al. 2002; Spahn et al. 2019), and even through activity like that of comets (e.g., Hsieh et al. 2004; Hsieh & Jewitt 2006; Jewitt et al. 2009; Jewitt 2012; Agarwal et al. 2017; Chandler et al. 2019).

Farther from the Sun, too, dust is abundant—from the dusty debris orbiting several of the solar system's small bodies (e.g., Stansberry et al. 2004; Braga-Ribas et al. 2014; Ortiz et al. 2015, 2017) to the ring systems around the giant planets (e.g., Elliot et al. 1977; Smith et al. 1979; Goldreich & Tremaine 1982; Smith et al. 1989). In short, dust is a pervasive presence in the solar system.

4.7.1. The Temporal Evolution of Dust in the Solar System—Steady State or Time-Variable?

4.7.2. Chondrites: A Window Into the Formation of the Solar System

Over the past decade, narrow ring systems and other orbiting material have been discovered orbiting several of the solar system's small bodies (as detailed below, in Table 3). Such gravitationally bound material is thought to be able to persist for long periods of time, even in the absence of cometary activity from the host object, and may take the form of orbiting fragments (essentially moonlets), arcs or rings.

Table 3. The Properties of the Rings Around 10199 Chariklo, 2060 Chiron and 136108 Haumea

Object	Orbital Radius (km)	Widths (km)	References
Chariklo	390.6 ± 3.3 (inner)	7.17 ± 0.14 (inner)	[1]
	404.8 ± 3.3 (outer)	${3.4}_{-1.4}^{+1.1}$ (outer)
Chiron	324	10	[2]
Haumea	2,287	70	[3]

References. [1] Braga-Ribas et al. (2014), [2] Ortiz et al. (2015), [3] Ortiz et al. (2017).

Download table as:  ASCIITypeset image

The first definite evidence for material orbiting distant solar system small bodies came with the discovery of bona-fide satellites, such as Charon, Nix, Hydra, Styx and Kerboros around Pluto. Given the suggested collisional origin of those satellite families, it seems likely that their formation was preceded by the existence of temporary ring systems orbiting their parent bodies, as the material ejected from those bodies accreted to form its new moons.

In the last decade, orbital fragments (moonlets) have been detected around the Centaurs 29P/Schwassmann–Wachmann 1 (Stansberry et al. 2004; Gunnarsson et al. 2008; Womack et al. 2017) and (60558) 174P/Echeclus (Rousselot 2008; Fernández 2009). More excitingly, however, occultation observations of the dwarf planet Haumea and the Centaurs 10199 Chariklo and 2060 Chiron have resulted in the detection of narrow ring systems in orbit around them—just as, in 1977, the rings of Uranus were discovered during observations of a stellar occultation by the planet from the Kuiper Airborne Observatory (e.g., Elliot et al. 1977).

The first set of rings to be discovered orbiting an object other than one of the giant planets were confirmed around the Centaur 10199 Chariklo (Braga-Ribas et al. 2014). That work made use of a network of telescopes distributed across South America to observe the occultation of a faint (R = 12.4 mag) star by the Centaur. Such occultation observations are a great boon to researchers, as they allow the true size and shape of an object such as Chariklo to be determined on the basis of the durations of the occultation as observed from a variety of locations, in addition to information on any locations from which the occultation does not occur. By piecing together the various occultation "chords" observed in this way, it is possible to reconstruct the projected size and shape of the object in question. In the case of Chariklo, the sites that observed the occultation recorded the occulted star dimming and brightening on two occasions before the true occultation began, then repeating that behavior after the conclusion of that main occultation event. On the basis of their observations, Braga-Ribas et al. found that Chariklo hosts two narrow, dense rings, with orbital radii 391 and 405 km and widths ∼7 km and ∼3 km, respectively.

Once the rings around Chariklo had been announced to the world, Ortiz et al. (2015) returned to earlier observations of the Centaur 2060 Chiron, which had exhibited a similar behavior during its own occultation of a faint background star. The observed pre- and post-occultation dimming of Chiron had originally been explained as being the result of jet-like features around the Centaur's nucleus (Ruprecht et al. 2013), which has long been known to exhibit cometary activity throughout the entirety of its orbit (e.g., Luu & Jewitt 1990; Meech & Belton 1990; Bus et al. 2001). Instead of being the result of such jetting, Ortiz et al. (2015) proposed that Chiron, too, hosts a dense, narrow ring, some 10 km wide, with an orbital radius of approximately 324 km.

More recently, observations of a stellar occultation by the dwarf planet 136108 Haumea, on 2017 January 21 revealed that, in addition to its satellites, Hi'iaka and Namaka, that dwarf planet, too, hosts a dense, narrow ring (Ortiz et al. 2017; Winter et al. 2019). The orbital radius of that ring is far wider than those around Chariklo and Chiron (which is not unexpected—indeed, the diameter of Haumea is so great that the rings of the two Centaurs would lie within its solid body), at ∼2287 km. Its width is also significantly larger than those around the two Centaurs, estimated at ∼70 km. Interestingly, the ring appears to be coplanar with the orbit of Haumea's outer moon, Hi'iaka, and the equatorial plane of the dwarf planet—suggesting a potential common origin for the moons and rings that is entirely in keeping with the idea that they were created in a significant impact event. Additional evidence for such an event can be found in the extended Haumea collisional family (e.g., Brown et al. 2007; Ragozzine & Brown 2007; Schlichting & Sari 2009; Leinhardt et al. 2010; Lykawka et al. 2012; Volk & Malhotra 2012; Vilenius et al. 2018), as discussed in Section 4.1.2.

The exact origin of the rings around the two Centaurs remains unclear, but theories include debris from cometary activity (Pan & Wu 2016; Wierzchos et al. 2017), a collision between the small body and another body, or between an orbiting satellite and another body (Melita et al. 2017), the tidal disruption of an orbiting satellite (El Moutamid et al. 2014), and the tidal disruption of the Centaur itself due to a close encounter with a planet (Hyodo et al. 2016). At this time, no single origin theory can be discounted, and observations of the two Centaurs only hint at possible origin scenarios. 2060 Chiron has been known to display cometary activity at heliocentric distances beyond the water-ice sublimation distance of approximately 3 au (Bus et al. 1991; Womack & Stern 1999; Jewitt 2009; Womack et al. 2017), which might well help to maintain its ring system. To date, 10199 Chariklo has not been observed to display such activity, but it is certainly possible that it may have done so in the relatively recent past (Wood et al. 2017). So, ring formation by cometary activity remains a possibility.

Given the frequency with which Centaurs are expected to experience close encounters with the giant planets, Wood et al. (2017, 2018) performed detailed n-body simulations to determine whether those encounters could be close enough to either disrupt the rings, or to help generate them (through tidal disruption of their parent bodies). Those simulations revealed that encounters that are close enough to create or disrupt rings around Centaurs should be very rare—sufficiently rare, in fact, that it seems unlikely that either Chariklo or Chiron would have experienced such an encounter during their time in the Centaur region. In other words, the results of those simulations argue against the idea that the rings of either body were created by a close encounter with a giant planet. Furthermore, the lack of disruptive encounters during the lifetimes of those Centaurs means that a primordial origin for the rings (or, at least, an origin that predates the transfer of their hosts to the Centaur region) cannot be ruled out.

In the coming years, it seems highly likely that further occultation observation programs will reveal additional rings orbiting more of the solar system's small bodies—indeed, given the scarcity of such observations for those small bodies in the past, the fact that three ring systems have already been discovered may argue that such rings are in fact relatively commonplace—revealing once again that our knowledge of the minutiae of the solar system's small body populations remains incomplete, and serving as a reminder that there is always something new to discover!

As was discussed earlier, the solar system is littered with debris moving on dynamically unstable orbits. In the absence of replenishment, these small body populations would quickly become depleted, on timescales of just a few million years, as a result of collisions with the solar system's planets (e.g., Zahnle & Mac Low 1994; Hammel et al. 1995; Sánchez-Lavega et al. 2010; Hueso et al. 2018) or the Sun (e.g., Michels et al. 1982; Farinella et al. 1994; Jones et al. 2018), fragmentation and disintegration (e.g., Crovisier et al. 1996; Sekanina 2000; Jenniskens & Lyytinen 2005; Jenniskens & Vaubaillon 2007), and ejection from the system as a result of close encounters with one or other of the system's giant planets (e.g., Levison & Duncan 1994, 1997; Bottke et al. 2002; Horner et al. 2004b).

Given the number of objects we observe in the various unstable populations, it is obvious that they must be continually replenished with fresh material, injected from those populations in the solar system that are nominally dynamically stable. In this manner, the near-Earth asteroids are held to be primarily sourced from the asteroid belt (e.g., Wetherill 1988; Morbidelli et al. 2002a; Morbidelli & Vokrouhlický 2003); the short-period comets from the trans-Neptunian region (e.g., Duncan & Levison 1997; Levison & Duncan 1997; Di Sisto & Brunini 2007; Lowry et al. 2008; Volk & Malhotra 2008, 2013), with contributions from the inner Oort cloud (e.g., Emel'yanenko et al. 2005; Brasser et al. 2012; de la Fuente Marcos & de la Fuente Marcos 2014b) and the Jovian and Neptunian Trojans (e.g., Horner & Lykawka 2010a; Horner et al. 2012c; Di Sisto et al. 2019; Holt et al. 2020); and the long-period comets from the Oort cloud (e.g., Öpik 1932; Oort 1950).

At the same time, vast quantities of dust are released into the solar system by cometary activity, and by collisional processes throughout the system—evidence of which can be seen on any clear night, in the form of meteors, the Zodiacal Light (e.g., Leinert 1975; Leinert et al. 1981; Reach et al. 2003), and the Gegenschein (e.g., Brandt & Hodge 1961; Wolstencroft 1967; Roosen 1970; Buffington et al. 2009; Ishiguru et al. 2013). Since dust is continually produced, the fact that the solar system is not substantially dustier is clearly the result of the efficient clearing of dust, by a variety of processes.

In this section, we describe the non-gravitational mechanisms by which planetary systems are kept clear of dust, and those that can help to facilitate the liberation of objects from the dynamically stable regions of the system, to join the various populations of dynamically unstable object.

4.9.1. Effects that Remove Dust

Three major processes dominate the lifetimes of dust grains in a planetary system; Poynting–Robertson drag, radiation pressure, and collisional destruction (e.g., Wyatt & Whipple 1950; Burns et al. 1979; Backman & Paresce 1993; Lagrange et al. 1995; Moro-Martín & Malhotra 2003; Krivov 2010; Wyatt et al. 2011). All of these processes act on timescales much shorter than the stellar lifetime, implying that the dust that permeates our solar system, and the debris discs observed around other stars, are not primordial (e.g., Backman & Paresce 1993; Wyatt 2008; Moro-Martín 2013). An unseen population of larger asteroidal and cometary bodies (planetesimals) are invoked as the mass reservoir for the ongoing production of dust grains we observe around other stars (Artymowicz & Clampin 1997; Dominik & Decin 2003). Analogously, the sources of dust in the solar system include the collisional grinding of asteroidal bodies and material ejected from active comets (e.g., Nesvorný et al. 2010; Koschny et al. 2019).

Radiation Pressure: The dynamics of small dust grains in the solar system are dictated by a combination of the incident stellar radiation field and the gravitational potential of the Sun. The strength of influence that radiation pressure has on a given dust grain can be described by the ratio of the radiation pressure force F_rad to the stellar gravitational force F_grav acting on it, defined as the β parameter (Burns et al. 1979; Krivov 2010). The β ratio is calculated as follows:

$$\begin{eqnarray*}&&\beta =\displaystyle \frac{{F}_{\mathrm{rad}}}{{F}_{\mathrm{grav}}}=\displaystyle \frac{3{L}_{* }{Q}_{\mathrm{pr}}}{16\pi {{cGM}}_{* }s\rho }.\end{eqnarray*}$$

Radiation pressure is proportional to the stellar luminosity, L_*, the grain cross-section, and the efficiency Q_pr with which momentum is transferred from photons to the dust, averaged over the stellar spectrum. The gravitational force, on the other hand, is proportional to the stellar gravitation parameter, GM_*, as well as the grain mass, and hence, to its volume and its density, ρ. Since both gravitation and radiation intensity are r⁻² laws, the value of β is dimensionless, and is fixed for a dust grain of given radius, s, and density, ρ. Dust grains effectively see a reduced gravitational potential from the central star due to this incident radiation. As a result, for dust grains released by objects⁵⁵ moving on circular orbits, or around the perihelion passage of objects moving on eccentric orbits, the influence of radiation pressure drives dust grains onto more eccentric orbits than those occupied by their parent bodies.

For sufficiently large and opaque grains, the radiation pressure efficiency, Q_pr, is constant near unity. As a result, the β ratio becomes inversely proportional to the grain radius, s. Radiation pressure is less and less important as grain size increases, becoming insufficient to perturb the orbital motion for particles larger than ∼100 μm (e.g., Pawellek et al. 2019). Equally, very small dust grains are inefficient absorbers and scatterers of incident stellar radiation. Hence, in the Rayleigh regime, where grains are much smaller than the dominant stellar photons, Q_pr is reduced, and becomes proportional to the grain radius, s. As a result, the β ratio levels off to a lower, constant value, at these small sizes. The exact size range over which radiation pressure is important depends on both the properties of the host star and the composition of the dust.

The grain size at which β = 0.5 is often called the blow-out limit. For dust originating from a larger parent body on a circular orbit moving at Keplerian velocity around the host star, grains with β values <0.5 will remain bound to the star. In such a scenario, grains for which β > 0.5 will become unbound from the host star upon release from their parent body. They are blown out of the system on hyperbolic orbits, leaving on orbital timescales, vacating the terrestrial region on a timescale of years, and leaving the Edgeworth–Kuiper Belt on a timescale of centuries. Around the Sun, the blow-out size of compact silicate dust is ≈1 μm, whereas around an A-type star the blow-out size of identical dust is ≈10 μm (Backman & Paresce 1993).

For dust originating from larger bodies moving on eccentric orbits, the limit between bound and unbound depends on the point in their parent body's orbit at which the dust grains are released (Kresák 1976):

$$\begin{eqnarray*}&&{\beta }_{\mathrm{lim}}=\displaystyle \frac{r}{2a}\end{eqnarray*}$$

where r is the distance at which the grain is released from its parent body, and a denotes the semimajor axis of that object. For grains released closer to perihelion, the excess kinetic energy results in a blow-out limit that falls in the range 0 < β_lim < 0.5 (dependent on the exact motion at the time). By contrast, for grains released closer to an object's aphelion, the limiting value for dust grain blow-out falls in the range 0.5 < β_lim < 1, and initially, for low β values, emission near aphelion will actually cause dust grains to move on orbits more circular than that of their parent (as seen in Figure 15, below). Indeed, for particles release at aphelion, a β value that is identical to the orbital eccentricity of the parent body will result in a dust grain moving on a circular orbit.

Zoom In Zoom Out Reset image size

For cometary parent bodies, where orbital eccentricities, e, are often very high, and dust production occurs primarily in the months around their perihelion passage, r_emission ≈ q = a (1 − e), and the limiting β ratio can be approximated with β_lim ≈ (1 − e)/2. For orbital eccentricities in excess of 0.99, which is typically the case for long-period comets, the lower size limit for bound grains can thus exceed the circular blowout limit by a factor ≳100. This distinction is critical when it comes to the discussion of dust behavior in other planetary systems. Studies of debris discs often discuss the "blowout limit" for the dust we observe in those systems—which brings with it an implicit assumption that the parent bodies are moving on circular, or near circular orbits. In reality, however, this blowout limit is actually an upper limit, since grains ejected by objects moving on eccentric orbits would be blown out of a system when identical grains emitted from a circular orbit would not be removed. This is particularly relevant when one considers observations of systems containing exo-comets (e.g., β Pictoris; Kiefer et al. 2014; Greaves et al. 2016). The stability and distribution of dust grains released from parent bodies on circular and eccentric orbits are illustrated in Figure 15.

It is challenging in the extreme to get an overview of the dust halo of the solar system because we are buried deep within it. As a result, while significant work has studied the evolution of dust in the context of meteor showers and the zodiacal light (as described in Section 4.7.1), the discovery of debris orbiting other stars has proved a great boon to researchers, since it allows the study of the entirety of the debris in a given planetary system at a single time. The presence of extended haloes of sub-blowout sized grains (the radii of which may extend up to several microns) has been observed in the optical and near-infrared scattered light images of debris disk-host stars (e.g., Smith & Terrile 1984; Schneider et al. 2014; Ren et al. 2019). For example, multi-wavelength observations of HD 181327 revealed a difference in debris ring location between optical and millimeter wavelengths that is consistent with radiation pressure induced dust grain size segregation. The millimeter wavelength observations also revealed a broad halo component (or unresolved second belt) to the disk (Marino et al. 2016).

It would be expected that any debris disk with a host star luminous enough to drive dust out of its system⁵⁶ should exhibit such a halo structure, and it has been suggested that up to half of the emission from debris discs at infrared wavelengths could be produced by sub-micron dust grains, for those stars luminous enough to drive the ejection of dust in this way (Thébault & Kral 2019).

We note here that not all dust haloes observed around other stars are the result of radiation pressure. For example, the young M-dwarf star AU Mic has a bright debris disk surrounded by a significant extended halo. However, as an M-dwarf, the star is clearly insufficiently luminous for that halo to be the result of radiation pressure driven escape. Instead, the presence of a halo is attributable to the strong stellar wind from the young M-dwarf star (Schüppler et al. 2015; Boccaletti et al. 2018). Similar to the pressure induced by stellar radiation, the pressure exerted by stellar wind particles is mostly directed radially away from the star, with the intensity dropping as r⁻². Stellar wind pressure is proportional to, and varies with, the stellar mass loss rate and the wind velocity. In the solar system, typical wind pressures are by more than three orders of magnitude lower than radiation pressure (e.g., Burns et al. 1979).

In combination with the influence of Poynting–Robertson drag (as described below) and stellar wind forces, radiation pressure induces a radial distribution of dust grains within a debris disk that is dependent on both the size of the particles and their (initial) orbits (e.g., Burns et al. 1979; Strubbe & Chiang 2006; Krivov 2010). While in the solar system, we can see individual objects as they contribute to the system's dust budget, determining the source of dust in extrasolar debris discs is more challenging. To accurately trace the location of the dust-producing planetesimals within those discs it is therefore necessary to observe larger dust grains which are less strongly influenced by these radiation forces. Modeling of debris discs from mid-infrared to millimeter wavelengths suggests that dust seen at wavelengths in the far-infrared and beyond traces the location of the underlying planetesimal belt (Pawellek et al. 2019).

Poynting–Robertson Drag:  The Poynting–Robertson effect (Poynting 1904; Robertson 1937) induces the in-spiralling of a dust grain orbiting the Sun through that grain's interaction with a "head wind" of solar photons. This head wind is the result of the orbital motion of the dust grains, relative to the mostly radial stellar radiation field. While a dust grain at rest would just be affected by the radiation pressure described above, any orbital motion leads to an additional force proportional to −v/c, where v denotes the grain's orbital velocity, and c is the speed of light. The momentum transfer that causes both radiation pressure and the Poynting–Robertson effect happens when the solar photons are absorbed or scattered by the grain. Subsequent re-emission, if anisotropic, can then lead to additional forces on macroscopic objects such as the Yarkovsky effect (which we describe in more detail in Section 4.9.2, below).

This perceived "head wind" causes a loss of orbital energy, resulting in the gradual decay of the dust grain's orbit. The Poynting–Robertson effect acts to reduce both the semimajor axis and the eccentricity of the dust grain's orbit. Grains therefore migrate toward the Sun from the initial orbits into which they were released until they eventually reach distances where the material sublimates. This happens at a few Solar radii for silicate dust and at several astronomical units for volatiles (e.g., Kobayashi et al. 2011). For grains moving on roughly circular orbits, the timescale for loss via Poynting–Robertson drag is given by

$$\begin{eqnarray*}&&{t}_{\mathrm{PR}}=\displaystyle \frac{{{ca}}^{2}}{\beta {{GM}}_{* }}=\displaystyle \frac{16\pi {c}^{2}s\rho {a}^{2}}{3{L}_{* }{Q}_{\mathrm{pr}}}.\end{eqnarray*}$$

Here, t_PR is a function of the grain radius, s, density, ρ, and the radiation pressure efficiency Q_pr, combined with the orbital radius of the grain, a, around a star with luminosity L_* (Wyatt & Whipple 1950; Burns et al. 1979). The timescale is proportional to grain size and makes Poynting–Robertson drag most effective on μm-sized dust grains, for which it causes grains to spiral inward from 1 au to a few Solar radii in roughly one thousand years. Smaller grains are quickly blown out of the system by direct radiation pressure. For dust grains initially moving in the region of the Edgeworth–Kuiper Belt, the timescale increases by a factor of approximately one thousand (to a million years, or so). Similarly, increasing the grain size to millimeters would again yield an increase of the timescale by a factor of a thousand.

While direct stellar wind pressure has little influence on dust dynamics and lifetimes in most systems, wind drag is more effective. The radial velocity of the wind particles (∼100 to 1000 km s⁻¹) is roughly three orders of magnitude smaller than that of the photons. The dust's azimuthal motion and the momentum transfer from the corresponding "head wind" resulting from the stellar wind particles becomes more significant; the additional factor c/v_wind makes wind drag comparable in efficiency to radiation-induced Poynting–Robertson drag. Gustafson (1994) estimates that wind drag contributes about one quarter of the total drag, at times of average Solar wind strength. Around stars with stronger stellar winds, wind drag can dominate (e.g., Reidemeister et al. 2011; Schüppler et al. 2015).

Collisional Destruction: The impact craters that cover most solar system objects bear witness to the importance and frequency of mutual collisions (as described in depth earlier in this manuscript). Relative speeds on the order of kilometers per second mean that such collisions often end the lives of dust grains. At the same time, collisions among bigger objects produce fresh dust. A cascade of collisions transfers mass from planetesimals down to observable dust. Once the material is ground down to sufficiently small sizes, radiation pressure and Poynting–Robertson drag move the dust grains out of the system or toward the Sun, respectively. Objects in regions and of sizes for which the loss timescale due to mutual collisions is shorter than the direct removal timescale due to radiation pressure and drag forces are called collisionally dominated bodies.

For a dust grain moving on a Keplerian orbit, its collisional lifetime, t_coll, will be broadly defined by its orbital period, t_orb, and the cross-section surface density, σ (cross section per disk surface area; also known as normal geometrical optical depth, τ), of potentially harmful colliders in the disk through which it moves (e.g., Burns et al. 1979; Backman & Paresce 1993; Wyatt & Dent 2002), such that:

$$\begin{eqnarray*}&&{t}_{\mathrm{coll}}=\displaystyle \frac{{t}_{\mathrm{orb}}}{2\pi \sigma }.\displaystyle \frac{2i}{{v}_{\mathrm{rel}}/{v}_{k}}\sim \displaystyle \frac{{t}_{\mathrm{orb}}}{2\pi \sigma }\propto \displaystyle \frac{{a}^{3/2}}{\sigma {M}_{* }^{1/2}}\end{eqnarray*}$$

where i is the orbital inclination, v_rel the typical relative velocity, and v_K = (GM_*/a)^0.5 the Keplerian speed. The fraction containing the inclination and the velocities is of the order of unity, and through substitution of the Keplerian speed and the circumference of the orbit (=2πa) for t_orb, we obtain the relation in terms of the stellar mass, M_*, orbital radius, a, and surface density, σ.

A collision is commonly considered catastrophic if the largest remaining fragment has at most half the mass of the colliders. The critical energy required to disrupt and disperse an object increases with increasing size. The critical specific energy (the energy per unit mass) decreases up to object radii ∼100 m, and then increases for larger objects which are strengthened by self-gravity (e.g., Benz & Asphaug 1999; Leinhardt & Stewart 2012). The critical impactor size required for a catastrophic collision also increases with object size, and as a result, the rate at which such collisions happen decreases. The lifetime of an object against collision increases with increasing grain size, albeit not as strongly as the Poynting–Robertson timescale. Below a critical grain size, Poynting–Robertson drag acts on shorter timescales than collisional disruption, while above that critical size, collisions dominate. As a result of the different dependencies in t_coll and t_PR, the critical size at which t_coll = t_PR is smaller in denser or more distant regions (e.g., Kuchner & Stark 2010). Observations of the interplanetary dust bands and counts of microcraters suggest that the critical size is around ∼100 μm (corresponding to ∼10⁻⁵ g) in the inner solar system (e.g., Fechtig & Grün 1975; Le Sergeant d'Hendecourt & Lamy 1980; Grün et al. 1985; Grogan et al. 2001).

While in the solar system, there is a clear distinction between dust grains large enough to be collisionally dominated and those small enough to be driven by radiation pressure and Poynting–Robertson drag, the same is not true of the debris we have observed beyond the solar system. With increasing disk surface density (or mass), the critical size at which t_coll = t_PR decreases until it becomes comparable to or smaller than the blowout limit. At that point, all dust is either bound and dominated by collisions or moving on unbound orbits. At the current epoch, our instrumentation is simply not sensitive enough to detect true analogues of the solar system's dust environment. As a result, all of the currently known debris discs are collisionally dominated, as a direct result of observational biases that favor the detection of bright, dense and cool dust discs (see e.g., Matthews et al. 2010, 2014; Hughes et al. 2018).

Detections of near-Edgeworth–Kuiper Belt analogues (∼10 × F_EKB) around Sun-like stars have been made at far-infrared wavelengths in continuum emission by Herschel (Eiroa et al. 2013; Sibthorpe et al. 2018), and have allowed us to begin to place our solar system's dust environment in the context of its peers. These detections approach the predicted limits on Edgeworth–Kuiper Belt flux of 1.2 mJy at 70 microns (Vitense et al. 2012) set by IRAS and COBE at 60 μm, favoring an upper limit <1 MJy sr⁻¹ (Aumann & Good 1990; Backman et al. 1995), and in situ dust measurements from New Horizons (Han et al. 2011). Those observations suggest that the Edgeworth–Kuiper Belt lies around the mid-point of dust disk brightness based on an extrapolation from the observed systems (Moro-Martín et al. 2015), but still lies within the collisionally dominated regime.

In the typical situation where a collisional cascade constantly replenishes and removes material, the size-frequency distribution of dust grains can be approximated by a power law, dN ∝ s^−Pds, with index p ≈ 3.5 (Dohnanyi 1969; O'Brien & Greenberg 2003), where dN/ds is the number of particles per unit grain radius. For the smaller grains that are produced in collisions but whose lifetimes are shortened by Poynting–Robertson drag, the size distribution flattens to an index in the range 2.5 < p < 3.0 (Reidemeister et al. 2011; Wyatt et al. 2011). Such flattening is observed for the dust distribution in the inner solar system (e.g., Fechtig & Grün 1975).

The combination of a collisional cascade and a sharp lower size cutoff set by radiation pressure creates a distinctive wavy pattern in the size distribution of dust grains, deviating from the single power law. The small grains that are unbound by radiation pressure are significantly underabundant with respect to those that are still small, but big enough to stay bound. The latter, barely bound grains, therefore lack a reservoir of potentially destructive colliders; their lifetimes and abundance are increased. In turn, yet bigger grains suffer more strongly from collisions with these barely bound grains; they are again underabundant, and so on (Campo Bagatín et al. 1994). The wavelength, amplitude, and grain size to which this pattern extends depend on the details of the collision physics, including material properties and impact velocities (e.g., Thébault et al. 2003; Krivov et al. 2006). The wavy size distribution can translate to observable waves in the spectral energy distribution (Thébault & Augereau 2007; Kim et al. 2018). When orbital eccentricities induced by radiation pressure exceed those inherited from the belt of parent bodies, collision velocities and rates are increased, the abundance decreased. For parent belts with very low dynamical excitation (e ≲ 0.01), this can cause an additional depletion of the barely bound grains up to grain radii s corresponding to β(s) ≳ e, resulting in a more tenuous halo (Thébault & Wu 2008).

In a collisional cascade, the destruction of two asteroidal sized bodies in a system produces large numbers of small dust grains, causing a temporary brightening of the observed debris disk. Such events occur stochastically, and the brightness increase is short lived (t_coll ∼ 10⁴ yr, Rieke et al. 2005; Jackson et al. 2014). The dust grains produced in such a collision are swiftly removed from the system by the subsequent collisional cascade and dynamical evolution through radiation forces as described above. The presence of very bright asteroid belt analogues orbiting other stars (with L_dust/L_* > 10⁻³, T_dust > 200 K; approximately ten thousand times more luminous than our own asteroid belt) is believed to be evidence of ongoing collisional processes, possibly analogous to the giant impacts seen in the solar system's youth (as described in Section 4.1; see e.g., Song et al. 2005; Fujiwara et al. 2009, 2010; Lisse et al. 2009; Melis et al. 2010; Kenyon & Bromley 2016). The rapid evolution of such bright asteroid belt analog debris discs, including both stochastic variability on periods of months to years and monotonic dissipation over timescales of years to decades, has been observed in a handful of systems (Melis et al. 2012; Meng et al. 2012, 2014, 2015, Su et al. 2019), and suggests that the violent processes that shaped our solar system are far from unique.

4.9.2. Effects that Remove Large Objects

The great majority of near-Earth asteroids are thought to be sourced from the asteroid belt. The primary routes by which these asteroids are transferred from the belt to the inner solar system feature interaction with mean-motion and secular resonances with the giant planets (e.g., Wetherill 1988; Morbidelli et al. 2002a). However, those mechanisms are so effective, on astronomical timescales, that they rapidly clear any debris from their regions of influence in the belt. To maintain a steady flux of new material to the inner solar system through those mechanisms, then, fresh material must first be transferred into those regions from elsewhere in the belt. To address this problem, Morbidelli & Vokrouhlický (2003) investigate the influence of a non-gravitational force known as the Yarkovsky effect on the main belt asteroids, finding that that effect could cause the migration of asteroids through the belt to the regions of instability that direct those asteroids into the inner solar system.

The diurnal Yarkovsky effect, explained in detail in Bottke et al. (2006), and the references therein, involves the absorption and re-emission of photons on an orbital body. Photons of light from the Sun hit an object and are absorbed. As the object is rotating, and its surface has a significant, non-zero, thermal inertia, the infrared radiation of the photon happens at a different angle to the absorption. This produces a small net force. In a prograde rotator, the force has a component in the orbital direction that creates a net increase in orbital rate (e.g., Bottke et al. 2006). The consequence of this small force is an increase in semimajor axis over time. For a retrograde rotator, the result is a decrease in orbital momentum, resulting in a decrease in semimajor axis. The magnitude of the force depends upon how close a body is to the Sun, the obliquity of the body's spin axis with respect to the orbital plane, and the body's thermophysical characteristics (e.g., Bottke et al. 2006). Mathematically, this is represented in equation:

$$\begin{eqnarray*}&&{\left(\displaystyle \frac{{da}}{{dt}}\right)}_{\mathrm{diurnal}}=-\displaystyle \frac{8}{9}\displaystyle \frac{\alpha {\rm{\Phi }}}{n}W\left({R}_{\omega }^{{\prime} },{{\rm{\Theta }}}_{\omega }\right)\cos \gamma \end{eqnarray*}$$

The radiative pressure coefficient (Φ) is dependent upon the radius of the object (R) and the Solar radiation flux (F).

$$\begin{eqnarray*}&&{\rm{\Phi }}=\displaystyle \frac{\pi {R}^{2}F}{({mc})}.\end{eqnarray*}$$

The Stefan–Boltzmann constant (α = 1 − A) comprises the bond albedo of the object (A). The thermal component, (W (Rω', Θω)) is dependent on the radius scaled for penetrative depth $({\text{}}R{\omega }^{{\prime} }={\text{}}R/\updownarrow v)$ and the thermal parameter (Θω). The thermal parameter (Θω) is to account for the thermal properties of the body.

The area to mass ratio of the object has a major effect on how the Yarkovsky effect influences an object. A large object has a large mass to area ratio, reducing the effect of such a small force. If an object is too small, the thermal gradient over the object lessens and the radiative difference becomes minimal. It is generally accepted that the size range where the diurnal Yarkovsky effect has the most influence is between one meter and approximately 10 km (e.g., Bottke et al. 2006).

In addition to the diurnal Yarkovsky effect, a seasonal effect has been described (Bottke et al. 2006). Unlike the diurnal Yarkovsky effect, the seasonal effect can only slow the object, reducing the semimajor axis.

$$\begin{eqnarray*}&&{\left(\displaystyle \frac{{da}}{{dt}}\right)}_{\mathrm{seasonal}}=-\displaystyle \frac{4}{9}\displaystyle \frac{\alpha {\rm{\Phi }}}{n}W\left({R}_{n},{{\rm{\Theta }}}_{n}\right){\sin }^{2}\gamma .\end{eqnarray*}$$

From the above equation, due to the radiative flux (F), it can be seen that the Yarkovsky effect has the most effect in the inner solar system. Simulations have shown that the Yarkovsky effect has an effect on the evolution of the near-Earth Asteroids (e.g., Morbidelli & Vokrouhlický 2003; Farnocchia et al. 2013), Martian Trojans (e.g., Ćuk et al. 2015) and the main asteroid belt (e.g., Nesvorný & Bottke 2004). Further out in the solar system, the impact of the Yarkovsky effect on the Jovian Trojans has been found to be negligible on objects larger than 100 m (Wang & Hou 2017), with objects more distant than Jupiter likely not experiencing the effect to any significant degree.

The Near-Earth asteroid population is of great interest as a result of the potential hazard it poses to the Earth (see Section 4.2.2), and the Yarkovsky effect will play an important role in altering the orbits of these objects, causing the degree to which they can be considered to be hazardous objects to change over time. As a result of the effect's influence on the orbital evolution of NEAs, there has recently been a concerted effort to detect the Yarkovsky effect on individual members of the NEA population, with over 40 confirmed detections to date (e.g., Vokrouhlicky et al. 2015; Ďurech et al. 2018), including the OSIRIS-REx target, (101955) Bennu (Deo & Kushvah 2017; Hergenrother et al. 2019).

Over the past two decades, the influence of the Yarkovsky effect on the migration of asteroids across the asteroid belt has been widely studied, and asteroidal drift under the influence of the Yarkovsky effect has become a key metric in the identification of asteroid collisional families, and the determination of their ages (e.g., Bottke et al. 2001; Bolin et al. 2017, 2018; Spoto et al. 2015; Delbó et al. 2019; Paolicchi et al. 2019).

Since the dawn of the Exoplanet Era, over 4000 exoplanets have been identified in approximately 3000 systems,⁶³ with many more candidate planets and systems discovered by NASA's TESS and Kepler missions awaiting confirmation using ground based facilities. The vast majority of these systems were identified through transit photometry, which looks for the almost imperceptible "wink" as an unseen companion passes across the face of the star as viewed from the solar system. The Kepler spacecraft (e.g., Borucki et al. 2010; Howell et al. 2014) proved to be the most productive planet search project to date, performing what was essentially the first census of the Exoplanet Era. Between its primary and K2 mission phases, Kepler led to the discovery of 2748 planets, with a further 3307 still awaiting confirmation.⁶⁴

Despite the apparent wealth of exoplanets discovered in the last thirty years, our current understanding of the exoplanetary systems we have discovered remains far less than our knowledge of the solar system's planets before the launch of the first interplanetary missions in the 1960s. Our knowledge, in the main, is limited to knowing the planet's minimum mass (if discovered using the radial velocity technique) or its size (using the transit method). A small fraction of the planets discovered to date have been observed directly—but even then, we can usually gain only limited understanding of the planet's nature (with the mass of such planets estimated based on an assumed age and their observed brightness; e.g., Marois et al. 2008).⁶⁵ Beyond these basic characteristics, little more is known for the vast majority of known exoplanets. In some cases, a bulk density can be calculated (for those planets that have been observed using both the transit and radial velocity techniques; e.g., Siverd et al. 2012; Masuda 2014; Johns et al. 2018, and many others). In systems containing multiple planets, some of which transit their hosts, it is possible to obtain improved mass and orbit measurements, and even to uncover the presence of additional planets, through the analysis of transit timing variations (TTVs)—which result from the effect of one planet's gravity perturbing the orbit of another (e.g., Agol et al. 2005; Holman & Murray 2005; Lithwick et al. 2012; Becker et al. 2015; Gajdoš et al. 2019; Hamann et al. 2019; Lam et al. 2020). Specific molecular species have been identified in the atmospheres of a small number of planets through multi-wavelength transit photometry (e.g., Charbonneau et al. 2002; Knutson et al. 2014; Khalafinejad et al. 2017) or direct measurement of the planet's spectrum (e.g., in the case of HR 8799 c; Oppenheimer et al. 2013; Wang et al. 2018a). But, in the main, our knowledge of the planet's physical nature is limited to either a minimum mass, or a size—nothing more.⁶⁶

Prior to the panoply of exoplanet discoveries beginning in the late 1980s, speculation on the types and abundance of planets within the galaxy was mostly inspired by planetary formation theory, which was rooted in the solar system. The Sun, an average star, hosts a set of planets that seemed to be loosely arranged in their orbits by size and mass, with rocky (terrestrial) inner planets, and more distant gaseous giants. It seemed reasonable to conclude that other Sun-like stars would like have similar groups of planets, arranged in similar ways, with rocky worlds interior to gaseous ones.

The first exoplanets to be discovered around Sun-like stars were far different to what was expected following this line of thought. 51 Pegasi b (Mayor & Queloz 1995) is comparable in mass to Jupiter, but orbits its host at a distance of just over 0.05 au, with a period of 4.23 days. In the years that followed, such "hot Jupiters" dominated exoplanet discoveries—a finding that led to significant innovation in our models of planet formation (e.g., Dawson & Johnson 2018, and references therein). The variety of exoplanets that have been found has revealed an unexpected diversity—in other words, not all planetary systems resemble our own.

With that said, as the sample size of known exoplanets has increased, this conclusions we had made on planet formation based on the sample of one system (the solar system) have somewhat stood the test of time. Our broad understanding (planet formation in a disk around a young star) has proven accurate, and we now know that planets are ubiquitous—most, if not all stars host planets. However, the sheer diversity of these planets—in size, mass, and orbital architecture—has proven to be far beyond the scope of solar system-based predictions for the demographics of exoplanets.

Figure 16 shows the dramatic rise in the number of known exoplanets, organized by detection technique. The two most productive methods are transit and radial velocity (RV) techniques. Here, we will summarise the current state of knowledge in exoplanet demographics as determined through transit and RV surveys. Figure 17 shows the distribution of known exoplanets as a function of their discovery method, their mass, and their orbital period. It is immediately apparent that the transit and RV methods of exoplanet detection are sensitive to different subsets of the population of exoplanets. Both techniques fall victim to significant observational biases, and are limited by the available observational temporal baseline, which effectively determines the maximum orbital period at which a given survey they can detect planets. The transit method is heavily biased toward detecting planets that move on short period orbits, and most effectively finds planets with larger radii (as they produce a significantly more obvious transit signature in the light curves of their host stars). Similarly, the radial velocity method is biased toward finding the most massive planets, with an efficiency that is limited toward short orbital periods by the available observational cadence, and toward long periods by the temporal baseline over which observations have been obtained.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

When interpreting the sample of known exoplanets, especially as displayed in plots such as Figure 17, it is critical to remember that what is known (and not yet known) about exoplanet demographics is shaped by the biases of the techniques used to detect exoplanets. These biases must be carefully accounted for to extract robust demographic information from the sample of detected exoplanets. To give just one example—when one examines Figure 17, it would be natural to think that hot Jupiters were abundant, and that most planetary systems feature at least one such scorching giant. This is, however, the result of the aforementioned observational biases. Simply put—both the transit and radial velocity techniques are strongly biased toward the detection of hot Jupiters—they are the easiest of all planets to find, and so one would expect them to dominate exoplanet demographics, particularly in the early days of the field. Indeed, the true occurrence rate of hot Jupiters is low—with studies based on data taken by Kepler and TESS consistently returning rates of less than 1% (e.g., Howard et al. 2012; Fressin et al. 2013; Wang et al. 2015; Zhou et al. 2019).

Numerous ground- and space-based transit surveys are responsible for having found most of the presently known exoplanets. One of the most notable efforts is the Kepler mission (Borucki et al. 2010), which acquired a nearly continuous four-year observational baseline for over 150,000 stars. Kepler identified over 4000 exoplanet candidates (Akeson et al. 2013; Thompson et al. 2018), of which more than 2300 have been confirmed or statistically validated (Rowe et al. 2014; Morton et al. 2016). Thanks to the observational baseline and photometric precision of Kepler, the orbital periods of these exoplanets range from less than a day to several hundred days and their radii are as small as ∼1 Earth radius (Twicken et al. 2016).

Hsu et al. (2019) provided a thorough analysis of the vetting and detection efficiency of the final data release for the primary Kepler mission (DR25) using updated stellar parameters from the Gaia DR2 release. They found that a typical FGK star hosts ${1.07}_{-0.08}^{+0.10}$ planets with radii in the range 1.0–16 R_Earth and orbital periods spanning 4–128 days. Hsu et al. (2019) also estimated that planets with sizes 0.75–1.5 R_Earth and orbital periods of 237–500 days have an occurrence rate for FGK stars of <27%. This value remains solely an upper limit, as a result of the low detection efficiency and high false alarm probability for planets in that region of parameter space. At even longer periods, an automated planet search through the Kepler data determined that, on average, Sun-like stars have 2.0 ± 0.7 planets between 0.1 and 1 Jupiter radii moving on orbits with periods between 2 and 25 yr (Foreman-Mackey et al. 2016). Interestingly, the exquisite precision of the Kepler photometry has enabled the investigation of specific sub-groups of the exoplanet population for which there are no analogues in the solar system. The radius distribution of short-period (<100 days) exoplanets contains a statistically significant decrease in occurrence between 1.5 and 2 Earth radii (Fulton et al. 2017). This feature is currently the subject of intense investigation, but is generally thought to be related to the erosion of planetary atmospheres by stellar radiation.

At longer orbital periods (hundreds of days), the reliability of candidates from transit surveys decreases and RV confirmation is typically employed to ensure the robustness of planet occurrence rate statistics. Santerne et al. (2016) combined Kepler photometry with long-baseline RV observations from the SOPHIE spectrograph to find that the total occurrence rate of giant planets (with transit depth between 0.4% and 3%) within an orbital period of 400 days is 4.6 ± 0.6%. At wider orbital separations, the efficiency of transit detections decreases sharply and long-baseline RV surveys provide much of the exoplanet demographic information. Between 0.3 and 7 au, giant exoplanets with mass in the range 0.3–13 Jupiter masses have an occurrence rate of ${6.2}_{-1.6}^{+2.8} \%$ (Wittenmyer et al. 2016b).

A primary goal of many studies has been to relate stellar properties to exoplanet occurrence rates. Since stars and planets form from the same protostellar material, it seems reasonable to expect that the properties of the host star would influence those of any planets. Such investigations were restricted to the realm of theory prior to the discovery of exoplanets, but now that planets have been discovered around a wide variety of stars, such studies are becoming ever more important.

In the early days of the exoplanet era, RV surveys of F-, G-, and K-type stars identified a positive correlation between the occurrence rate of giant planets within several au and stellar metallicity (Santos et al. 2003; Fischer & Valenti 2005). This is now commonly known as the "planet metallicity relation" for exoplanets, and is interpreted as providing strong support for the core accretion theory of giant planet formation (Pollack et al. 1996). Subsequent work has found that this relation likely extends to later-type (lower mass) stars (Johnson et al. 2010; Neves et al. 2013).

Planet detection efforts aimed specifically at M dwarf stars have also uncovered interesting relationships between exoplanet demographics and stellar properties. In general, small planets are found to be much more prevalent around M dwarfs. According to Dressing & Charbonneau (2015) and Gaidos et al. (2016), the average M dwarf has roughly 2.2–2.5 exoplanets with radii between Earth and Neptune moving on orbits with periods of ∼200 days or less. M dwarfs also favor the formation of "compact multiple" systems of exoplanets, featuring several Earth-size planets on tightly packed short-period orbits, with the scale of the entire planetary systems comparable to that of the Galilean moons orbiting Jupiter (e.g., Muirhead et al. 2015; Ballard & Johnson 2016). While low mass planets orbiting late-type stars seem to be abundant, the prevalence of giant planets orbiting such stars is notably smaller than that observed around more Sun-like FGK stars (e.g., Cumming et al. 2008). In other words, it seems that the smaller the star, the smaller will be its largest planets.

As noted in Section 5.1, the architecture of the solar system dominated theories regarding formation processes and orbital evolution until the discovery of other systems. Over time, it has become clear that the range of possible architectures, although not a continuum, is incredibly vast (Winn & Fabrycky 2015). One way to place the solar system within the broader context is to consider the appearance and planetary signatures of our solar system if it were to be observed as an exoplanet system.

In Table 4, we show the semimajor axes and orbital periods for the planets in the solar system. Also included are the radial velocity semi-amplitudes (m s⁻¹) for the planets, under the assumption that the solar system is viewed close to edge-on. The typical precision of current RV surveys is ∼1 m s⁻¹, but next generation instruments are poised to be deployed (Plavchan et al. 2015). The "NN-explore Exoplanet Investigations with Doppler spectroscopy" (NEID) instrument aims to have total per measurement error budget of ∼30 cm s⁻¹ (Halverson et al. 2016). The "Echelle SPectrograph for Rocky Exoplanets and Stable Spectroscopic Observations" (ESPRESSO) instrument is anticipated to achieve a per measurement precision of ∼10 cm s⁻¹ (Pepe et al. 2014). Thus, although current RV surveys are only sensitive to Jupiter and Saturn analogues (subject to their having a sufficiently long observational baseline), Earth/Venus analogues may be detectable in the coming years. Note that although the semi-amplitude of Uranus and Neptune are large compared to the terrestrial planets, their exceptionally long orbital periods mean that the detection of their analogues continues to pose a significant challenge.

Also shown in the above table are the predicted transit depths (in parts per million, ppm) and transit probabilities for the eight solar system planets. The transit depth of Mercury lies at the very threshold of the photometric precision of the Kepler mission (Borucki et al. 2010), and while Kepler did detect planets with diameters smaller than or comparable to Mercury (e.g., Kepler-37 b Barclay et al. 2010; Kepler-444 b Campante et al. 2015), those planets moved on orbits with periods far shorter than that of Mercury (∼13.4 and ∼3.6 days, respectively). Such short orbital periods allowed large numbers of transits to be observed during Kepler's primary mission, which greatly increased the ease with which those planets could be detected.

However, the major challenge in the detection of solar system analogues lies in the transit probability, which follows an inverse relationship with the semimajor axis (Kane & von Braun 2008). Simply put, the more distant the planet from its host, the lower the likelihood of it being observed to transit from any given external location. Depending on the duration and window function of the survey (von Braun et al. 2009), most of the solar system's planets would, at best, be detected as single transit events by current and past surveys. Such single transit events are far from optimal, and pose significant challenges with regards to their follow-up and validation (Osborn et al. 2016; Villanueva et al. 2019; Yao et al. 2019). As a result, the bias of survey completeness toward short orbital periods means that the full context of true solar system analogues is an on-going process that will rely on combinations of exoplanet detection techniques to fully realize.

A further critical aspect of our solar system as an analog for exoplanetary systems is the realization that in situ data for exoplanets is unlikely to be achieved for several generations (and possibly significantly longer). Hence, all inferences regarding the connection between planetary atmospheric data and surface processes relies upon the in situ data obtained within the solar system from which models applicable to exoplanets may be constructed. The primary terrestrial sources of such atmospheric data and geological profiles are Venus, Earth, Mars and several of the moons of the giant planets, most particularly Titan.

Furthermore, it is important to study the evolution of the solar system's planets through time, since they occupy a specific evolutionary state at the present epoch (Del Genio et al. 2019). The most relevant solar system bodies to the study of terrestrial planets are Venus and the Earth since they represent the primary target demographic of exoplanet searches in the context of habitability (e.g., Kane et al. 2019b). To illustrate the importance of the study of our own planets, we note that the abundance of oxygen in the Earth's atmosphere was significantly lower than the present day for most of our planet's history (Lyons & Reinhard 2009), while recent studies have revealed that Venus could have retained its liquid surface water up until ∼1 Gya (Way et al. 2016). Searches for potential Venus analogues will play an important role in understanding the bifurcation of atmospheric evolution between Venus and Earth in our solar system (Kane et al. 2013, 2014). More detailed studies of the Venusian atmosphere are required in order to fully characterize runaway greenhouse environments for terrestrial exoplanets (Schaefer & Fegley 2011; Ehrenreich et al. 2012; Kane et al. 2018). Likewise, the solar system's giant planets are critically important laboratories for deriving atmospheric and interior models of giant exoplanets as a function of composition, age, and insolation flux (Mayorga et al. 2016; Withers & Vogt 2017; López-Puertas et al. 2018).

As mentioned previously, analogues of the solar system's ice giants are particularly challenging for us to detect at present. Despite this, data from studies of the solar system's ice giants are being utilized for exoplanet models (Arridge et al. 2014), and scaled down versions of the solar system may make the detection of ice giant planets beyond the snow line more accessible (Kane 2011). Overall, the expanding exploitation of the synergies between planetary science and exoplanetary science is an essential step in reliably inferring the derived properties of exoplanets, most particularly inferences regarding the surface conditions of terrestrial exoplanets.

Over the course of its history, the solar system has exhibited large variability both in the structures within the system and the visible emission from the attendant planets and debris belts that would be detectable by an outside observer. As a result, how the system would have looked to an external observer has changed dramatically over the past 4.5 billion years.

At the earliest epoch, around 3–10 Myr after the Sun arrived on the main sequence, the gaseous proto-planetary disk would still be in the final stages of dissipation, or have already dispersed. At that time, the outer solar system would be dominated by the giant planets, which would still be warm from accreting their gaseous envelopes, and would therefore be radiating brightly in the near infra-red (much as is the case with the four giant planets orbiting HR8799 e.g., Marois et al. 2008, 2010). At this stage the primordial debris belts would most likely still be dynamically cold, producing little dust through collisions, and therefore be faint. In the inner solar system, the terrestrial planets would not yet be fully formed, and the inner regions of the solar system would therefore be full of small, short-lived dust particles, as a chaotic pinball of proto-planetary embryos collided and aggregated to one day form the four terrestrial planets. At infrared wavelengths, the warm dust produced in this manner would be detectable as a bright infrared excess at near and mid-infrared wavelengths, perhaps with silica emission features as a result of the colliding embryos producing high temperature, gaseous SiO, as seen around HD 15407 (Fujiwara et al. 2012) and HD 172555 (Lisse et al. 2009; Johnson et al. 2012).

In the Nice Model for the evolution of the solar system, it was suggested that the migration of Saturn and Jupiter resulted in the two planets crossing a mutual mean motion resonance approximately 700 Myr after the solar system formed. The result of that resonance crossing was a massive dynamical instability, which disrupted the orbits of the giant planets and the solar system's debris belts. The orbits of the outer four planets become chaotic, and the quiescent primordial debris belts were shattered, scattering cometary and asteroidal material far beyond their point of formation by the rearrangement of the giant planets onto their current orbital configuration. The model suggests that this rearrangement resulted in the Late Heavy Bombardment, delivering a drenching of water rich cometary material to the terrestrial planets. Some of the scattered small bodies were captured into the Jovian and Neptunian Trojan populations (e.g., Morbidelli et al. 2005, Lykawka & Horner 2010), while the migration of Jupiter sculpted the Asteroid belt, gutting the inner system and removing ∼90% of its original mass. If the solar system did indeed experience a period of dynamical instability such as that proposed in the Nice Model, then the collisions driven by this proposed period of extreme instability would have resulted in the production of a large density of small dust grains, which would in turn produce an appreciable, although short lived (<10⁴ yr), brightening of the Sun's debris disk by several orders of magnitude (Wyatt et al. 2007; Booth et al. 2009; Raymond et al. 2011). Dust-producing collisions occur stochastically over the lifetime of a planetary system, such that the levels of dust (and hence observed infra-red excess) in the system will peak and decay many times.

Currently, at an age of 4.5 Gyr, the Sun's two principal debris belts are very faint, with fractional luminosities, L_dust/L_star ∼ 10⁻⁷ for the inner solar system (Backman et al. 1995—the definition of "Zodi") and ∼1.2 × 10⁻⁷ for the Edgeworth–Kuiper Belt (Vitense et al. 2012).⁶⁷ The peak of the continuum emission from the Edgeworth–Kuiper Belt (the solar system's cold disk) is 0.5 mJy at a wavelength of approximately 40–50 μm, as seen from a distance of 10 pc—a flux that is less than 1% of the contribution from the Solar photosphere at those wavelengths (Vitense et al. 2012). By contrast, the peak of the continuum emission from the zodiacal light falls at approximately 20 μm (Reach et al. 2003), at a flux level of 0.1 mJy, as seen from a distance of 10 pc (compared to a stellar flux at 2 Jy at the same wavelength).⁶⁸

These values are lower than those observed for the debris discs detected around nearby Solar type stars (Montesinos et al. 2016; Sibthorpe et al. 2018), which typically exhibit fractional luminosities for cold debris discs of the order 10⁻⁴–10⁻⁵ and warm discs (e.g., Fujiwara et al. 2013; Kennedy & Wyatt 2013) up to levels of a few percent of the stellar flux (e.g., BD+20 307, Zuckerman et al. 2008; HD 15407, Fujiwara et al. 2012). The known menagerie of debris discs around other stars therefore represent brighter, and more massive (and/or more collisionally active), analogues to the solar system's Edgeworth–Kuiper or Asteroid belts.

Current detection of circumstellar debris dust is limited by instrument sensitivity, and many fainter discs remain to be detected—the Edgeworth–Kuiper Belt is believed to have a disk brightness about average for its age (Moro-Martín et al. 2015), but is as yet undetectable. With this caveat, far-infrared observations by Herschel identified cool debris discs around ∼15% of nearby, Sun-like stars (Trilling et al. 2008; Eiroa et al. 2013; Marshall et al. 2014b; Montesinos et al. 2016; Sibthorpe et al. 2018), with a higher incidence of ∼30% around more massive A-type stars (Chen et al. 2006; Thureau et al. 2014).

The combination of volume-limited far-infrared surveys with exoplanet survey results for the same sample of stars makes a direct statistical comparison of the properties of stars with planets (down to a few Earth masses) and/or debris discs (with dust brightness levels ∼10× the Edgeworth–Kuiper Belt) possible for the first time. This led to the determination that stars with detected planets have an enhanced incidence of circumstellar debris (Matthews et al. 2014). Searching for more refined correlations, such as with planetary mass or stellar metallicity, has led to the identification of weak trends (Wyatt et al. 2012; Marshall et al. 2014a; Meshkat et al. 2017), but such analyses are hampered by the heterogeneity of the underlying disk and planet data sets, making direct determination of the impact of such properties on the occurrence of planets/debris difficult (Moro-Martín et al. 2015; Wittenmyer & Marshall 2015).

Spatially resolved imaging of the scattered light and/or the continuum emission from circumstellar debris has revealed the architectures of these planetary systems in greater detail. In this regard the HST, Gemini/GPI, and VLT/SPHERE at optical/near-infrared wavelengths, and ALMA operating at radio wavelengths have been revolutionary. Most obviously, the complex dynamical structure exhibited by β Pictoris' debris disk, believed to be induced by its directly imaged Jovian-mass companion, clearly exhibits the interplay between the planetary and planetesimal components in a system (e.g., Smith & Terrile 1984; Golimowski et al. 2006; Dent et al. 2014; Matrà et al. 2017, 2019).

At millimeter wavelengths, other systems have exhibited an equally intriguing variety of structures thought to be the result of disk-planet interaction, such as eccentric debris rings (MacGregor et al. 2017; Faramaz et al. 2019), gaps in broad planetesimal belts (Marino et al. 2018, 2019), multiple planetesimal rings (MacGregor et al. 2019), haloes of millimeter-sized dust grains from dust filtering (Marino et al. 2017; MacGregor et al. 2018) and scattered planetesimal populations (Marino et al. 2017; Geiler et al. 2019). At optical and near-infrared wavelengths, high angular resolution coronagraphic imaging reveals similar structures to those seen at millimeter wavelengths, including the presence of haloes of small dust, asymmetries, and eccentric rings (Schneider et al. 2014).

Initial successful efforts toward scattered light imaging of debris discs (building on the detection of the disk around β Pictoris; Smith & Terrile 1984) were dominated by the HST (Schneider et al. 2014; Choquet et al. 2018; Ren et al. 2019) and the reanalysis of archival observations (Choquet et al. 2016). More recently, advances in high contrast adaptive optics have led to a renaissance in ground-based disk imaging, with high profile results including the polarimetric imaging of HR 4796A (Perrin et al. 2015), the resolution of many young discs (e.g., Hung et al. 2015; Millar-Blanchaer et al. 2016; Chauvin et al. 2018; Hom et al. 2020) including new detections (Sissa et al. 2018), the identification of asymmetries (Draper et al. 2016) and complex substructures (Olofsson et al. 2018). The information gleaned from the dust distribution and properties in these two wavelength regimes is complementary, providing constraints on the dust grain scattering albedo and size (e.g., Choquet et al. 2018; Marshall et al. 2018).

Determining the distribution and intensity of the scattered light levels around other stars has been of particular interest due to its implications for future exoplanet direct imaging instruments (e.g., Roberge et al. 2012). This exo-zodiacal light would constitute a confusing background to any attempts at the direct imaging of exoplanets in the habitable zone of their host stars. The direct measurement (by photometry or spectroscopy) of exo-zodiacal light would be stymied by the brightness of the star. Interferometric methods have therefore been employed at optical and infrared wavelengths to infer exo-zodiacal dust levels through the measurement of the visibility deficit revealing the presence of an extended component (i.e., circumstellar dust) to the total emission from a stellar system.

Using the Keck interferometer, limits at the level of 100× to 1000× Zodi were obtained for tens of stars, with a few detections for the brightest discs (Mennesson et al. 2014). Measurements with VLTI surveyed around 90 stars at near-infrared wavelengths spanning a broad range of stellar spectral types, finding a hot dust incidence comparable to that of cool debris discs detected by Spitzer and Herschel but uncorrelated with the presence of a known outer debris belt (Absil et al. 2013; Ertel et al. 2014). Most recently, the Hunt for Observable Signatures of Terrestrial Systems on the Large Binocular Telescope Interferometer searched for dust close to the habitable zone of 38 stars, finding that the presence of habitable zone dust was correlated with cool dust, and that most Solar-type stars were not very dusty, with a median of 3 Zodis and a 95% upper limit of 27 Zodis based on their N-band measurements (Kennedy et al. 2015; Ertel et al. 2018, 2020).

Almost all stars begin their existence with a circumstellar disk most commonly identified through excess emission at infrared wavelengths (Williams & Cieza 2011; Ribas et al. 2014; Wyatt et al. 2017). The effects of grain growth (at the earliest stages) and collisional evolution lead to a predictable monotonic decay in disk brightness over time, assuming a steady-state evolution (e.g., Wyatt et al. 2007; Kains et al. 2011; Marshall et al. 2011; Löhne et al. 2012; Sierchio et al. 2014). The dust masses in circumstellar discs can be traced from the sub-millimeter emission, revealing a drop in the observed mass at around 1–10 Myr in age (e.g., Panić et al. 2013; Holland et al. 2017), comparable to the approximate lifetime of protoplanetary discs as seen through observation of accretion and gas tracers (Ribas et al. 2014; Richert et al. 2018). More massive and compact (smaller semimajor axis) discs evolve more quickly due to the shorter collisional timescales for dust-producing planetesimals, leading to the same overall brightness decay over time, which follows an approximate 1/t relation. A bright disk is not necessarily a massive disk,⁶⁹ except potentially at the earliest phase of its evolution.

While the evolutionary trend with time holds for the vast majority of debris discs, there exist several well-known anomalies, such as BD+20 307 (Song et al. 2005) and q¹ Eri (Liseau et al. 2008, 2010), that are brighter (higher fractional luminosity) than expected given the stellar age (see Figure 18). The interpretation of these abnormally bright discs is that the observed emission is, wholly or partially, the result of stochastic event(s) that have caused an enhanced rate of dust production at the present epoch, rather than being the result of those discs being unusually massive.

Zoom In Zoom Out Reset image size

The mostly likely mechanism behind this enhanced dust production is thought to be either self-stirring by the largest planetesimals within the disk (e.g., Kenyon & Bromley 2008; Krivov & Booth 2018), or the migration of a planet through the planetesimal belt (e.g., Mustill & Wyatt 2009). The process by which self-stirring could result in excess dust production is subject to many assumptions (such as the size and strength of the planetesimals in the disk), leading to orders-of-magnitude uncertainty in the timescale for the mechanism to act. At the most extreme end of systems with enhanced collision rates are bright (L_d/L_* > 0.01) exo-Asteroid belts that display evidence of rapid time variability or decay at near- or mid-infrared wavelengths (e.g., Melis et al. 2012; Meng et al. 2015), ruling out an in situ, steady-state origin for the observed dust. At far-infrared wavelengths, the detection of spatially resolved emission from overly bright debris discs around young stars (e.g., Moór et al. 2015, 2017a, 2017b, Vican et al. 2016); Vican et al. 2016) places a constraint on the disk evolution, revealing several systems that are consistent with either self-stirring or planet migration influencing the observed disk architecture. Given that these discs are anomalously bright, the stirring event should have only been a (relatively) recent occurrence in these systems.

For example, a recent detection of CO gas in the debris disk around HD 32297 reveals that the observed debris disk could be coeval with the gas disk, which in turn suggests that it might therefore be only a few thousand years old (Cataldi et al. 2020). This bright debris disk system is young, but otherwise unremarkable in its extent or architecture. As a result, this discovery suggests that the tacit assumption that the discs we observe around other stars are the same age as the host star and present as a result of (broadly) steady state evolution is not always robust.

A handful of debris discs have mid-infrared spectra with features that reveal the composition of their constituent dust. As a result, the origin of that dust, and the nature of its production, can be directly determined. The debris disk orbiting HD 172555 has SiO gas in its spectrum, consistent with high temperature, shocked material produced in a hypervelocity impact (Johnson et al. 2012), whereas that around HD 109085 reveals a mixture of primordial (icy, amorphous) and processed (crystalline) material consistent with an origin of the parent bodies in the outer belt of that system (Lisse et al. 2013).

As we become better able to study the structure and distribution of dust in exoplanetary systems, it will become possible to model the processes that cause that structure to occur. For example, the debris dust orbiting HR 8799 has been found to lie in two distinct reservoirs, one interior to, and the other exterior to, the star's four known giant planets. Much as is the case in the solar system, the structure of debris belts in the HR 8799 system is clearly driven by the gravitational sculpting resulting from the influence of those four planets (e.g., Su et al. 2009; Contro et al. 2016; Goździewski & Migaszewski 2018; Geiler et al. 2019). By contrast, the sharp inner edges to the dust distributions of two young stars (HR 4796A and HD 141569) have been proposed as being the result of the photophoretic driving of dust through a particularly tenuous dispersing gas disk (e.g., Krauss & Wurm 2005, Krauss et al. 2007; Hermann & Krivov 2007)—a process that has also been invoked to explain the origin of high-temperature inclusions in the solar system's asteroidal and cometary bodies (e.g., Mousis et al. 2007a, 2007b, Moudens et al. 2011).

Among known debris discs, Poynting–Robertson drag (as discussed in Section 4.9.1) has a minimal effect on the radial distribution of dust grains (Wyatt 2005, 2006). These systems are dense enough (∼10 to 1000 times the mass of the Edgeworth–Kuiper Belt) that the dust lifetime is dominated by mutual collisions rather than by migration. For fainter discs, Poynting–Robertson drag can be an important dust removal mechanism, but such systems lie beyond our present capacity to detect. The proposed Spica infrared/sub-millimeter mission, scheduled for launch and operation in the mid-2030's, would provide observations of many exact Edgeworth–Kuiper Belt analogues (Roelfsema et al. 2018).

In the coming years, new facilities will enable us to identify the origin, composition, and dynamics of the dust-producing bodies that comprise debris discs with greater precision than has been achieved to date. Such data will not only help us to better understand and characterize the debris discs in question—it will also play an important role in our efforts to address the more esoteric questions of habitability in these planetary systems in the longer term, since it will offer the potential for us to quantify the impact regimes that would be experienced by potential rocky habitable zone planets, following methods initially developed to study the evolution and history of our own planetary system (e.g., Horner & Jones 2008a, 2009, 2010b, 2010a).

A primary scientific aim of exoplanet searches is to determine the degree to which our solar system represents a typical outcome of planetary system formation and evolution. To this end, much attention and effort is lavished on the quest to detect Earth-like planets in Earth-like orbits. Indeed, this was the chief aim of the Kepler spacecraft mission (Borucki et al. 2010). Space-based transit surveys such as Kepler, CoRoT, and TESS (Barge et al. 2008; Ricker et al. 2015) are particularly adept at detecting such small planets. The photometric sensitivity of these missions reaches about 30 parts per million (ppm) for suitably quiet stars, while the flux decrement due to an Earth-size planet transiting a Solar-type star is 80 ppm. In the four years of the Kepler prime mission, a handful of potential Earth- analog candidates were identified moving on orbits within their host stars' habitable zones (Kane et al. 2016). While the radial velocity (RV) method has been a successful technique for detecting exoplanets for decades (Campbell et al. 1988; Cochran & Hatzes 1994; Butler et al. 1996), the RV signal of an Earth-mass planet in a 1 au orbit is a dauntingly diminutive 8 cm s⁻¹, which remains below the current best precision of a few tens of cm s⁻¹ for instruments such as HARPS and ESPRESSO (Fischer et al. 2016).

Of equal importance, but perhaps capturing less public interest, is the presence of outer giant planets analogous to Jupiter and Saturn, moving on orbits with periods of more than 10 yr. For these planets, the transit method is of minimal use, both because the probability of a planet transiting its host scales inversely to the planet's semimajor axis, and because the transit of such a distant planet would be a singular ∼1 day event in any given 10+ year period. Fortunately, however, the RV method is eminently capable of detecting Jupiter analogues: our Jupiter imposes an RV signal of 12 m s⁻¹ in a 12 yr period on the Sun, and RV surveys have achieved precisions of, or better than, 3 m s⁻¹ for at least the last 25 yr (Fischer et al. 2014; Endl et al. 2016). From the long-running RV surveys, there is a growing consensus that about 3%–6% of Solar-type stars host a "Jupiter analog," typically defined as being a giant planet orbiting beyond the ∼3 au snow line (Wittenmyer et al. 2006, 2011b, 2016b; Zechmeister et al. 2013; Rowan et al. 2016). As the temporal baseline of legacy RV surveys lengthens, true Saturn analogues may be detectable; the RV signal of Saturn is 3 m s⁻¹ in an almost 30 yr orbit. Given that the orbital periods of the ice giants Uranus and Neptune are 84 and 165 yr, respectively, and that their maximum RV amplitudes would be just 0.3 and 0.28 m s⁻¹, the RV method is not, currently, likely to deliver true analogues for those planets any time soon, and so to find such planets, we must turn to other detection techniques.

Direct imaging probes a different but complementary region of parameter space to the RV and transit methods, being most sensitive to giant planets moving on orbits at tens of au from their host stars. To date, the favorable contrast ratios of young planetary systems have biased direct imaging detections toward systems younger than ∼100 Myr. The HR 8799 system (Marois et al. 2008, 2010; Wang et al. 2018a) has been considered a benchmark "scaled up solar system analog," with four giant planets of mass 5–7 Jupiter masses orbiting between 14 and 68 au, as well as two distinct debris belts (potentially analogues of the Asteroid and Edgeworth–Kuiper belts; e.g., Su et al. 2009; Contro et al. 2016; Goździewski & Migaszewski 2018). As mentioned in Section 5.2, scaled down versions of the solar system may also assist in the detection of ice giant analogues via RVs and transits (Kane 2011).

As we noted earlier, the combination of transit and RV observations enables the density of certain exoplanets to be determined. In the main, these results have been used to draw conclusions on the potential composition of the planets in question, but it has also recently been used to highlight the significant disparity between the densities of two almost identically sized planets in the Kepler-107 system. Both Kepler-107 b and c have radii around 1.5–1.6 times that of the Earth, and so it would be natural to assume that RV observations would show them to have similar masses. However, such observations instead reveal that the bulk density of Kepler-107 c is more than twice that of Kepler-107 b (12.6 g cm⁻³ versus 5.3 g cm⁻³). As discussed by Bonomo et al. (2019), the best explanation for this disparity is that Kepler-107 c was once the victim of a giant, mantle-stripping collision—as was the case for Mercury in the solar system (e.g., Benz et al. 2007; see also Section 4.1.1, above). These results demonstrate that we are now entering an era where mass and size measurements of exoplanets can begin to give information not only on the composition of those planets, but also on their collisional history. Given the widespread evidence of giant impacts in the solar system, this result is perhaps not particularly surprising—but it is worth stressing that, without the evidence from our own backyard, planets such as Kepler-107 c would probably be hugely confusing additions to the exoplanet catalog. Given the uncertainties inherent in calculating the densities of transiting exoplanets (particularly the smallest), it seems unlikely that we would be able to definitively identify the Earth and Mercury as having relatively high densities compared to the other terrestrial worlds, were we observing the solar system from the outside. However, in the coming years, the precision with which we can make such measurements will improve as new facilities come on-line. In the future, we may well therefore reach a point where such forensic studies can be carried out for planetary systems similar to our own.

Currently there are in excess of two hundred known satellites orbiting the eight planets within the solar system (as detailed in Section 2, in Table 1), most of which orbit the two largest planets in our system, with Jupiter hosting 79 known moons and Saturn hosting 82 known moons. While none of those moons are large enough that we would be able to confidently detect them were they placed in orbit about the known exoplanets, the sheer number of moons in the solar system, particularly the large number orbiting the Jovian planets, indicate a high probability that moons will also exist orbiting exoplanets. As a result, there is a general consensus in the scientific community that exomoons do exist (e.g., Williams et al. 1997; Kipping et al. 2009; Heller 2012; Heller & Pudritz 2015; Zollinger et al. 2017). However, while recent work suggested the first preliminary detections of exomoon signatures (Teachey et al. 2018; Teachey & Kipping 2018), those results remain heavily debated, and as a result, no exomoons have been confirmed to date (Heller et al. 2019; Kreidberg et al. 2019).

The lack of confirmed detections should not, however, be taken as evidence that such moons do not exist. Instead, it is, at least in part, the result of such moons remaining beyond the limits of our detection ability, and it is likely that future surveys with more powerful tools will confirm that, like planets, exomoons are ubiquitous. It may, however, be that we have been looking in the wrong places. Astronomers speculate that exomoons are likely ubiquitous based primarily on our knowledge of the solar system's outer planets, yet, to date, we have not looked for moons around planets that are true analogues to the solar system's gas and ice giants. Despite extensive efforts to detect moons around exoplanets with orbital distances between 0.1 and 1 au (e.g., Kipping et al. 2012), the wait for a confirmed detection goes on. This dearth can potentially be explained by the instability of exomoons in systems that experience intense stellar tides (e.g., Barnes & O'Brien 2002) or inward migration (Spalding et al. 2016). Long-period transiting exoplanets may indeed be better candidates for studies of exomoons than their short-period counterparts.

One of the key challenges inherent to the detection of exomoons is the impact of the orbital motion of the moon around the planet as the planet orbits its star. It has been suggested that a potential method by which exomoons could be detected is through their effect on the transits of their host planet across its star. There are two ways in which such variations could manifest and be detected—through transit duration variations and/or TTVs (Brown et al. 2001; Kipping et al. 2012, 2013). In addition, as the planet transits its host star, the moon may also be seen to transit. However, as the planet orbits the star, the moon also orbits the planet, and so the position of the moon in its orbit about the planet will change each time the planet transits the star. The signature of the moon will therefore vary its position within the transit of the planet depending on its orbital phase, as shown in Figure 19, below. The problems caused by this variability in the timing of the moon's transit relative to that of its host planet are further exacerbated by the possibility that there may be occasions when the moon is lined up perfectly with its host planet, and therefore transiting or passing behind the planet as the planet is in transit. Such a scenario will mean that the transit signature of the moon would disappear for that particular transit, leaving a signal that would be indistinguishable from a solitary, moonless planet.

Zoom In Zoom Out Reset image size

The time at which the planet transits can also be affected by the existence of an exomoon. The moon–planet system orbits about the center of mass of the two objects (their barycenter). In the absence of any perturbations from other planets in the system, the barycenter of the planet–moon system will follow a Keplerian orbit around the star. At those times when the moon trails the planet, the planet must therefore be leading the barycenter in its orbit around the star (since the planet and moon must always be directly opposite one another across the barycenter). As a result, when the moon trails the planet, the planet will be seen to transit its host star a little earlier than would otherwise be expected. Equally, at times when the moon is leading the planet in their orbit around their host star, the planet must therefore lag behind, trailing the barycenter. As a result, the planet will be observed to transit the star later than would otherwise be expected. Thus the variance of the transit timing of the planet can be an indication of the existence of an exomoon. However this method is also used to determine the existence of additional planets in orbit around the star causing a similar gravitational effect on the transiting planet. As a result, caution is needed when using this method for exomoons. Other potential methods that have been suggested for exomoon detection include microlensing (Han & Han 2002; Bennett et al. 2014), pulsar timing variation (Lewis et al. 2008), radio emission modulations (Noyola et al. 2014) and direct imaging (Cabrera & Schneider 2007). Hill et al. (2018) found that instruments would need the capacity to resolve a separation of 10–4400 μ arcseconds in order to directly image and resolve a planet–moon system. Direct imaging could also reveal the existence of a moon through the transit of a moon or of its shadow across a bright planet (Heller 2016).

The current focus of exomoon searches on planets at relatively small orbit radii is the direct result of the fact that the majority of the most promising techniques for the detection of exomoons rely on the detection of transits of the moon's host planet. Such transit observations are strongly biased toward short period planets, with transits growing more unlikely as the orbital radius of the planet increases. Two giant exoplanets moving on orbits well beyond 1 au that are possible candidates for in-transit exomoon detection are Kepler-421b (Kipping et al. 2014) and Kepler-167e (Kipping et al. 2016). The ephemerides of these planets have been found to be free of TTVs (Dalba & Muirhead 2016; Dalba & Tamburo 2019), enabling the accurate prediction of future transits that may be searched for photometric signals of exomoons.

Over the past few years, the study of exomoons has become a topic of great interest, with particular focus on the possibility that such moons might increase our chances of finding potentially habitable worlds beyond the solar system. Studies on the occurrence rates of giant planets have found that they are less likely to be found in the habitable zone of their star than smaller terrestrial planets. However, if each giant planet hosts more than one moon, then habitable zone exomoons may well be more numerous than in the habitable zone terrestrial planets (Hill et al. 2018).

One reason that exomoons are considered particularly interesting in the context of the search for life beyond Earth is that they would likely offer a variety of sources of energy to a potential biosphere, rather than relying on the flux received from the host star—an idea fueled by our observations of Io and Europa in the Jovian system. The reflected light and emitted heat from the host planet could help to ensure a clement climate on such a moon, while tidal heating could drive volcanic activity, providing both additional energy and nutrients to any life that developed there (Heller & Barnes 2013; Hinkel & Kane 2013). These combined heating effects act to increase the size of the habitable zone around the host star of such moons, creating a wider temperate area in which they could maintain conditions that are amenable to liquid surface water (Scharf 2006). At the same time, occultations of the host star by the giant planet could help to reduce the incoming energy flux for satellites moving close to the inner edge of the habitable zone—preventing such a moon from overheating. This interplay between tidal and radiative heating from the host planet and the diminution of the incoming stellar flux during occultation events is potentially complex, with the inclination of the moon's orbit, and the presence of additional moons in the system likely to play an important role. For an exomoon orbiting an exoplanetary Uranus, for example, occultation events would occur seasonally, with the satellite spending most of its time in full starlight. At the same time, the orbital eccentricity of the planet itself will doubtless impact upon the potential habitability of its moons. For a planet like BD+14 4559 (whose orbit is shown below, in Figure 20) that spends part of its orbit outside the outer edge of the habitable zone, the extra tidal and radiative energy delivered by the host planet could potentially enable any potentially habitable exomoon to maintain its habitability during this time, though it should also be noted that such a moon would likely lack the time to freeze over during these excursions beyond the nominal outer edge of the habitable zone.

Zoom In Zoom Out Reset image size

In addition, as we have seen in our own planetary system, large exomoons offer the possibility for liquid water to be found far beyond the boundary of the habitable zone, at least as it has traditionally been defined. Moons like Europa, Enceladus, and Titan have become targets of great interest for astrobiological research, despite the fact that they lie many au beyond the outer edge of the solar system's habitable zone. While such moons would clearly be of great interest were they found orbiting Jupiter-analog exoplanets, it is likely that they would be poor targets for the search for life elsewhere. While it is reasonable to assume that Europa is a habitable world, for example, it is not fair to consider that it would be detectably habitable. Europa's ocean is buried beneath a kilometers thick ice sheet—and even in our own backyard, we cannot yet say whether that ocean contains life. For this reason, the focus on exomoon habitability, in a detectable/measurable sense, will remain on those satellites of giant planets that themselves orbit within (or close to) the habitable zone of their host star—such satellites would offer not only the possibility of being considered potentially habitable worlds, but also the possibility that their habitability could be remotely probed, using the next generation of astronomical telescopes and instrumentation.

In the solar system, Earth and Neptune bound a clear division between the terrestrial and gaseous planets. Separated in size by a factor of four, no known solar system object has a radius larger than Earth's but smaller than Neptune's. It was natural, therefore, to expect that exoplanetary systems might follow this same behavior—rocky, terrestrial worlds close to their host stars would be relatively small (as in the solar system), while the more distant giant planets would all be large—Neptune-sized or bigger.

It has become obvious, however, that this is not the case. One of the great surprises to come out of the Kepler mission was that, rather than being rare, planets between the size of Earth and Neptune are actually abundant. Kepler found such planets in droves—as can be seen in Figure 21, below. Now, it is widely known that super-Earths—likely terrestrial worlds with radii slightly larger than the Earth—and sub-Neptunes—likely planets with thick gaseous atmospheres, reminiscent of the solar system's ice giants, with radii slightly smaller than Neptune—are by far the most common of the planets that we can currently detect in the galaxy. Some authors have even suggested that such planets might be the most abundant of all—outnumbering their smaller and larger siblings—but at the same time, logic would dictate that there are likely to be more small planets than large ones in the cosmos (just as, in the solar system, there are more small bodies than larger ones, and just as there are more small, low mass stars than massive stars). The true size distribution of planets smaller than the super-Earths remains to be uncovered in the coming decade—but regardless of the degree to which they are the most common planets in the cosmos, the super-Earths and sub-Neptunes pose a fascinating conundrum, showing where our ability to infer planetary properties based on the solar system breaks down.

Zoom In Zoom Out Reset image size

Prior to the era of Kepler, the surprising abundance of super-Earth and mini-Neptune exoplanets was already becoming clear, with the detection of such planets being achieved by several RV surveys. For planets moving on orbits with periods less than fifty days, it is clear that smaller planets are more common than giants—with the occurrence rate of exoplanets on such orbits rising by approximately a factor of ten as one moves from Jupiter-mass objects to those only a few times more massive than the Earth (Mayor et al. 2009; Howard et al. 2010; Wittenmyer et al. 2011a). The sensitivity of Kepler to exoplanets in this region of parameter space enabled a more detailed investigation of super-Earth and mini-Neptune prevalence. Accounting for both false positives and completeness, Fressin et al. (2013) found the occurrence rates of both super-Earths and mini-Neptunes with periods less than 85 days to be ∼23%, higher than any other size planet in any region of parameter space considered in their study.

Understanding the occurrence rate of super-Earths and mini-Neptunes is of critical importance to theories of planet formation. Early population synthesis models that simulated planet formation via core accretion anticipated an increasing planet occurrence rate with decreasing planet mass for objects orbiting at several au (e.g., Ida & Lin 2004; Alibert et al. 2005; Mordasini et al. 2009). However, the observed prevalence of super-Earth and mini-Neptune exoplanets at orbital distances within 1 au challenged the theoretical models. During formation, planets whose mass grew to several times that of Earth were expected to either spiral into the host star via Type-II migration or rapidly accrete a gaseous envelope to become a gas giant planet (Ida & Lin 2008; Schlaufman et al. 2009; Alibert et al. 2011)—a process of runaway growth that should result in planets rapidly passing through the super-Earth/mini-Neptune mass range. Producing a super-Earth or mini-Neptune planet would therefore require the unlikely dispersal of the gas disk at a particular time, truncating that process of runaway growth. The mere existence of so many super-Earth and mini-Neptune exoplanets in the galaxy demonstrated that this theory was flawed, and has led to numerous suggested mechanisms to explain how these exoplanets may have formed (e.g., Ida & Lin 2010; Rogers et al. 2011; Hansen & Murray 2012; Chiang & Laughlin 2013; Cossou et al. 2014).

Super-Earths and mini-Neptunes occupy a fascinating region of the mass–radius parameter space whereby vastly different interior and atmospheric compositions are a priori plausible for a given planet. These planets can have substantial fractions of silicate rocks, metals, ices, and gases (e.g., Valencia et al. 2007; Rogers & Seager 2010). A subset of super-Earth and mini-Neptune exoplanets that are particularly amenable to transit and RV characterization have had their radii and masses measured precisely (e.g., Marcy et al. 2014; Weiss & Marcy 2014). Substantial effort has been made to explore the atmospheres and interiors of this subset of well characterized exoplanets. Lopez & Fortney (2014) used planetary evolution models to suggest that, for mini-Neptunes, the planet's radius can be used as a proxy for its hydrogen-helium envelope fraction. Wolfgang et al. (2016) employed a probabilistic mass–radius relationship model to explore the intrinsic astrophysical scatter present in the relation between super-Earth and mini-Neptune mass and radius, while Rogers (2015) demonstrated that most 1.6 Earth-radius exoplanets have densities that are too low to be explained by just iron and rock, suggesting that these exoplanets are not terrestrial in nature. Beyond the small group of studies listed here, the atmospheres and interiors of these transitionary worlds are the subject of intense ongoing investigation.

Given the ubiquity of super-Earth and mini-Neptune exoplanets in the galaxy, one must wonder: why is the solar system devoid of this class of planet? In other planetary systems, the presence of cold, long-period giant exoplanets has been found to be strongly correlated to the existence of inner super-Earth-sized companions (e.g., Zhu & Wu 2018; Bryan et al. 2019). One potential explanation for the lack of such a planet in the inner solar system is that Jupiter could have acted as a barrier to Uranus, Neptune, and Saturn, halting their inward migration and preventing them from becoming super-Earth or mini-Neptune class planets (Izidoro et al. 2015a). Equally, given proper disk conditions, super-Earths may have existed in the solar system previously, clearing the material inside of Mercury's orbit before colliding with the Sun (Martin & Livio 2016). Yet another explanation could be that the solar system does contain a super-Earth or mini-Neptune planet that has so far eluded detection (see Batygin et al. 2019 for a review). Additional explorations of the small and large bodies nearby and across the galaxy will likely shed light on this peculiar aspect of the solar system.

Exoplanetary atmospheres are a window to an exoplanet's composition, formation history, evolution, and even habitability. The proximity of the solar system planets has afforded us the luxury of being able to study their atmosphere up-close and personal. Unlike exoplanets, which cover just a single pixel through the largest telescopes, the solar system's planets are easily resolved, allowing the behavior of their atmospheres to be studied in detail. The advent of planetary space exploration, in the 1960s, meant that we could get closer still—sending orbiters to observe the planets continuously, and even launching probes that could dive into their atmospheres, and, for the terrestrial planets, land on their surfaces. This has allowed us to investigate the atmospheres of the planets in our solar system to a level that will most likely not be possible for an exoplanet in the foreseeable future. Almost all of the information we can gather pertaining to an exoplanet's surface, formation, and evolution will be filtered through distant observations of its atmosphere.

Not long after the discovery of the first exoplanet came the first theoretical predictions regarding exoplanet atmospheric characterization. Seager & Sasselov (2000) proposed that measurements of the wavelength dependence of an exoplanet's transit depth would reveal variations in atmospheric opacity. This could, in turn, be used to infer the properties and composition of an exoplanetary atmosphere through a technique known as transmission spectroscopy. This technique was first successfully used to identify sodium in the atmosphere of HD 209458 b based on four transits observed by the HST (Charbonneau et al. 2002).

Since the initial characterization of the atmosphere of HD 209458 b, several other techniques for exoplanet atmospheric characterization have been developed. The observation of the occultation of short-period, highly irradiated exoplanets by their host stars (secondary eclipses⁷⁰) led to the first detection of photons directly emitted by an exoplanet atmosphere (Charbonneau et al. 2005; Deming et al. 2005). The observation of secondary eclipses enables the calculation of planetary Bond albedo and atmospheric temperature. Similarly, Knutson et al. (2007) conducted infrared observations of half of the orbital phase of the hot Jupiter HD 189733 b, effectively mapping the temperature distribution across planetary longitude. Phase curve maps, such as those presented in that work, have provided valuable information about energy transport and heat redistribution in the atmospheres of the hottest exoplanets.

Exoplanet atmospheres are also being characterized through high dispersion ground-based spectroscopy, whereby spectral lines of an exoplanet are detected separately from stellar lines thanks to the time varying radial motion of the exoplanet (e.g., Snellen et al. 2010). This technique has lead to confident detections of carbon monoxide and water vapor in exoplanet atmospheres (e.g., Brogi et al. 2012; Birkby et al. 2013), has enabled the estimation of day–night wind velocities on exoplanets, and has provided validation for 3D exoplanet atmosphere models (Kempton et al. 2014).

At greater orbital separations, direct imaging has also become a viable technique for the atmospheric characterization of young, self-luminous exoplanets. The HR 8799 system contains three directly imaged exoplanets that have been observed for over a decade, revealing critical information about the atmospheres and orbital dynamics of young giant planets (e.g., Marois et al. 2008, 2010). However, the immense planet–star contrast and the fundamental limits on angular resolution set by diffraction add difficulty to the direct observation of smaller or cooler (i.e., more mature) exoplanets at visible and near-infrared wavelengths.

Among the various methods of exoplanet atmospheric characterization, transmission spectroscopy is presently the most productive means of probing exoplanet atmospheres. Although the entire sample of exoplanets that have been subject to transmission spectroscopy observations are hotter and have shorter-period orbits than any object in the solar system, substantial efforts have been made to simulate the transmission spectra of the solar system's planets. For Venus, Barstow et al. (2016) generated synthetic spectra using a radiative transfer model to determine that the cloudiness of a Venus-analog exoplanet would preclude atmospheric characterization from an observatory such as the James Webb Space Telescope (JWST), even if the hypothetical planet orbited a small star only 10 parsecs away. Pondering the Earth as an exoplanet has become a cottage industry within exoplanetary science. With appropriately high-signal observations, transmission spectroscopy of an Earth-analog exoplanet could reveal the presence of key biosignature gases as well as other species such as ionized calcium (e.g., Pallé et al. 2009). Importantly, mid-transit observations of an Earth-twin would be hindered by atmospheric refraction, which would mask atmospheric information below ∼10 km (e.g., García Muñoz et al. 2012).

Transmission spectra have also been synthesized for Jupiter, Saturn, and Titan. Irwin et al. (2014) used spectroscopic observations and a radiative transfer model to simulate the transmission spectrum of Jupiter between 0.4 and 15 μm. The transmission spectrum of Jupiter as measured in a single transit would yield accurate constraints on the vertical profiles of temperature and hydrocarbon abundance in Jupiter's stratosphere (Irwin et al. 2014). Dalba et al. (2015) demonstrated that occultations of the Sun by Saturn observed by the Cassini Spacecraft in the near-infrared could be used to derive a transmission spectrum between 1 and 5 μm. The spectrum displayed multiple absorption features due to methane, as well as features from the hydrocarbon byproducts of Saturn's photochemistry (i.e., acetylene, ethane, and higher-order aliphatic hydrocarbons). Saturn's transmission spectrum also displayed a flat continuum between 1 and 5 μm caused by atmospheric refraction. Titan's transmission spectrum was also reconstructed from Cassini solar occultation observations (Robinson et al. 2014). It displayed absorption features from methane, acetylene, ethane, higher-order hydrocarbons, and potentially carbon monoxide. Multiple scattering by the thick haze present in Titan's atmosphere gives the spectrum a distinctive slope increasing toward shorter wavelengths, which is comparable in size to the strong absorption features (Robinson et al. 2014).

In order to fully understand the formation and evolution of our solar system, a variety of computational tools have been developed that allow the dynamical evolution of both the small bodies and the planets themselves to be studied on billion year timescales. These n-body dynamics packages, such as SWIFT (Levison & Duncan 1994), Mercury (Chambers 1999), and REBOUND (Rein & Liu 2012) can clearly also be used to study the dynamics of exoplanetary systems. They can be used, for example, to study the impact regimes that might be experienced by Earth-like planets in those systems (e.g., Horner & Jones 2008a, 2009, 2012; Horner et al. 2010), or to assess which of the known exoplanetary systems could host potentially habitable planets, based on the gravitational influence of the known planets in the system (e.g., Jones et al. 2005, 2006; Agnew et al. 2017, 2019, 2018a, 2018b; Horner et al. 2020).

A more immediate and important use for such tools in the study of exoplanetary systems is to use them to assess whether the proposed planets within a given system are actually dynamically feasible. As the number of known exoplanets has increased, the rigor with which new discoveries are investigated, prior to being announced, has fallen. For the first few discoveries, researchers and referees worked to the maxim that "extraordinary claims require extraordinary evidence," and all possible alternatives to the planetary hypothesis were examined in great detail. In recent years, as the discovery of exoplanets has become routine, an ever increasing population of planets moving on unusual orbits, or orbiting peculiar stars, have been announced (e.g., Wolszczan & Frail 1992; Lee et al. 2009; Beuermann et al. 2010, 2013; Potter et al. 2011; Ramm et al. 2016).

Given the importance of new exoplanet discoveries in helping theorists to better understand planetary formation, it is clearly vital to ensure that each newly proposed planetary system is what it seems to be. Since the great majority of such systems are found by indirect means, there is clearly a risk that some other physical process might be producing a variation that could be misinterpreted as the influence of a planet orbiting a given star. Indeed, the detection of exoplanets is hindered by a plethora of phenomena that can cause stars to vary in a fashion similar to that expected to be caused by planetary companions (such as sunspots, magnetic cycles, periodic variability, companion stars, and many others; e.g., Dumusque et al. 2011; Robertson et al. 2014; Meunier et al. 2015; Fischer et al. 2016; Rajpaul et al. 2016; Nicholson et al. 2018). It is therefore useful to have a "sanity check" for each newly proposed planetary system, in order to make sure that the candidate planets are reasonable.

Dynamical studies of newly discovered planetary systems therefore provide an excellent tool by which the veracity of those systems can be examined. If the candidate planets prove to be dynamically stable on timescales comparable to the lifetime of the system, then they can be considered to be dynamically feasible. On the other hand, if dynamical study of a given system shows that the planets therein interact so strongly that they collide, or eject one another from their host system on timescales far shorter than the system's lifetime, this either suggests that we have fortuitously observed the system during its death throes (a statistically unlikely but certainly plausible possibility) or that the proposed planets do not exist, at least on the orbits claimed.

When such simulations are carried out, the results fall into three broad categories. The first are those systems for which no dynamical instability is observed across the full spread of architectures tested (e.g., Wittenmyer et al. 2014b). These systems prove so dynamically stable that it is perfectly reasonable to assume that the planets truly exist on the orbits that are proposed. Beyond this, however, such results show the systems to be so dynamically stable that there is likely to be plenty of space left for further planets in that system—in other words, they indicate systems where further observations may well prove fruitful in coming years.

The second class of exoplanetary system are those for which such simulations reveal only narrow islands of stability across the phase space studied. Examples of the dynamical maps for two such systems are shown in Figure 22, below. In both cases, the candidate planets move on orbits that are close to mutual mean-motion resonances. In the case of HD 200964 (the left-hand plot; Wittenmyer et al. 2012b), our simulations show that the orbits of the two candidate planets are only stable when they are trapped in mutual 4:3 mean-motion resonance (the narrow island of stability visible at the center of the dynamical map). Similarly, for HD 204313 (the right-hand plot; Robertson et al. 2012b), the planets are only dynamically stable when trapped in mutual 3:2 mean-motion resonance. In both cases, as soon as the planets are moved away from a mutually resonant solution, they begin to strongly perturb with one another, resulting in one or other of them being flung into the center star, the two planets colliding with each other, or one being ejected from the system entirely.

Zoom In Zoom Out Reset image size

As a result, we can safely conclude that the planetary systems proposed are dynamically feasible provided that the planets therein are trapped in mutual mean-motion resonance. Such results have an interesting additional outcome—not only do they demonstrate that the systems can be dynamically feasible, such simulations have enabled significant additional constraints to be placed on the orbits of the planets in a number of systems (e.g., Robertson et al. 2012a, 2012b; Wittenmyer et al. 2014a, 2016a), above and beyond those that result purely from the observational data.

The third class of system are those for which the proposed planets in a given system move on orbits that appear highly dynamically unfeasible—sometimes even crossing one another's path (e.g., Horner et al. 2011, 2012d, 2013, 2014b, 2019; Wittenmyer et al. 2012a; 2013; Hinse et al. 2014a, 2014b; Marshall et al. 2020). While it is certainly feasible for mutually crossing orbits to be dynamically stable (examples in our own solar system include the Jovian and Neptunian Trojans, and the Plutinos; e.g., Morbidelli et al. 2005; Di Sisto et al. 2010, 2019; Horner & Lykawka 2010a; Pirani et al. 2019), such stability is typically facilitated by mutual mean motion resonance between the orbits of the objects in question. Even for newly discovered objects in our own solar system, it is critically important to confirm whether their orbits are dynamically stable or dynamically unstable (e.g., Lykawka & Mukai 2007; Gladman et al. 2008; Horner & Lykawka 2010a; Horner et al. 2012c; 2011a; Di Sisto et al. 2014, 2019; Wu et al. 2018). For exoplanetary systems, such dynamical studies are even more important—since strong dynamical instability would infer that the planetary system, as proposed, is simply not feasible, and that either the system is dramatically different to that proposed in the literature, or the observed variations that led to the announcement of planets around a given star must have another explanation.

In recent years, a number of multiple planetary systems have been proposed orbiting very tightly bound eclipsing binary stars,including several around what are known as "post-common envelope binaries" (hereafter PCEBs). Those systems consist of a highly evolved primary star (typically, though not always, a white dwarf) with a secondary stellar companion (often a dim red dwarf star, similar to our Sun's nearest stellar neighbor, Proxima Centauri) orbiting at a remarkably small distance. The two stars, which orbit about their common center of gravity with periods of a couple of hours, move on orbits that are aligned such that they eclipse one another each time the secondary passes between the primary star and the Earth, along our line of sight. Multiple massive companions have been claimed orbiting several such binaries (e.g., Lee et al. 2009; Beuermann et al. 2010; Potter et al. 2011; Qian et al. 2011) on the basis of variations in the timing of the eclipses between them, with those timing variations proposed to result from the small wobble back and forth of the binary stars under the gravitational influence of distant, unseen companions. Of those systems, however, only two have stood up to dynamical scrutiny—namely, the planets orbiting NN Serpentis (e.g., Beuermann et al. 2010; Horner et al. 2012d; Beuermann et al. 2013; Marsh et al. 2014) and UZ Fornacis (Potter et al. 2011; Horner et al. 2014a). Even in the case of NN Serpentis, however, Mustill et al. (2013) showed that the planets could not have survived the evolution of the binary from its main-sequence to post-main sequence state, leading to the suggestion that, if the planets truly exist, they most likely represent a second-generation of planet formation, which occurred after the ejection of the star's common envelope, at the end of the white dwarf star's evolution away from the main sequence—a finding supported by Völschow et al. (2014).

In Figure 23, we show the best-fit proposed orbital solutions (top) and resulting dynamical stability plots (bottom) for the planets proposed to orbit the eclipsing binary stars QS Virginis (Lee et al. 2009; Horner et al. 2012a; left) and HW Virginis (Almeida & Jablonski 2011; Horner et al. 2013; right). In both cases, the proposed solutions prove to be catastrophically dynamically unstable across the whole range of possible solutions—with instability occurring on timescales of just a few hundred or few thousand years. Such extreme instability suggests that the origin of the observed variations in the timing of eclipses in these systems can not be planets, at least, not planets moving on the orbits proposed by Lee et al. (2009) and Almeida & Jablonski (2011). Indeed, given the fact that the planets proposed in the majority of such systems have failed to stand up to dynamical scrutiny, it seems likely that a common non-planetary explanation must be sought to explain the periodic variability in their eclipse timings—a possibility that researchers are continuing to explore as more data becomes available on these fascinating systems (e.g., Goździewski et al. 2015; Völschow et al. 2016; Nasiroglu et al. 2017; Navarrete et al. 2018, 2020).

Zoom In Zoom Out Reset image size

Regardless of the true origin of the observed eclipse-timing variations for these fascinating objects, the results described above demonstrate conclusively the importance of applying dynamical models, first developed to study the evolution of the solar system, to the orbital stability of newly discovered exoplanetary systems. In some cases, such simulations will simply reveal that a proposed planetary system is dynamically stable—but in other cases, they might reveal a false positive, or highlight a case where the process of fitting to the observational data has converged on a solution that does not represent the true global minimum of the orbital phase space that the planets could occupy. We therefore contend that such modeling should form a critical component of the planet discovery process, and have incorporated such analysis into the standard procedure of the Anglo-Australian Planet Search (AAPS) (e.g., Wittenmyer et al. 2014a, 2014b, 2017a).

While the orbital solutions of the transiting and radial velocity exoplanets are well constrained due to the requirement of confirming the cyclical nature of the signal before announcement of detection (along with short orbital periods in the case of transiting planets/hot Jupiters), identification of planets through multi-epoch direct imaging of exoplanet systems (e.g., HR 8799; Marois et al. 2008, 2010), looking for co-moving companions to the star, provide only weak constraints on the orbits of the planets as they are, by and large, on wide orbits, with decades-long orbital periods and therefore do not move much in year-to-year imaging. The dynamical stability of HR 8799 has been extensively studied (e.g., Marois et al. 2008; Goździewski & Migaszewski 2009, 2014, 2018; Marshall et al. 2010; Wang et al. 2018c) leading to a variety of more-or-less stable orbital solutions depending on the assumptions made by the authors of the orbital eccentricities of the planets and the inclination of the system. One consistent result is that three of the planets are trapped in a 4:2:1 mean-motion resonance (a scenario often known as a Laplace resonance, and one that mirrors the orbital architecture of Jupiter's moons Io, Europa and Ganymede). Unstable orbital solutions might then be indicative that the system is undergoing a dynamical rearrangement, analogous to the Late Heavy Bombardment in the solar system, rather than suggesting the planetary system is somehow spurious—a result that would be in keeping with the bright debris discs observed in the system (e.g., Su et al. 2009; Hughes et al. 2011; Matthews et al. 2014; Contro et al. 2016; Goździewski & Migaszewski 2018; Wilner et al. 2018; Geiler et al. 2019). In the future, as more multi-planet systems are discovered using techniques such as direct imaging and astrometry, dynamical methods will become even more important in determining which of the many plausible orbital architectures that could fit the observed data are most likely to be a fair representation of the true state of the newly discovered systems.

In the first few years of the exoplanet era, the search for alien worlds was dominated by radial velocity surveys that were limited in the number of stars they could observe, and the frequency with which they could carry out their observations (e.g., Latham et al. 1989; Cochran & Hatzes 1994; Walker et al. 1995; Baranne et al. 1996; Butler et al. 1996; Tinney et al. 2002; Naef et al. 2004). While those surveys were able to build up lengthy temporal baselines, and to begin finding ever more distant planets (e.g., Gregory & Fischer 2010; Ségransan et al. 2011; Vogt et al. 2017; Wittenmyer et al. 2017a; Kane et al. 2019a; Rickman et al. 2019a), the fact that they could only target a small number of stars limited the number of planets that they could discover.

In the middle of the first decade of the 21st Century, the first exoplanet transit surveys came online. Unlike the radial velocity surveys, these dedicated programs were able to observe all night, every night. By using wide-field cameras, those transit surveys (such as the Wide-Angle Search for Planets, WASP; e.g., Kane et al. 2004, 2005a, 2005b; Christian et al. 2006; Butters et al. 2010; HAT-Net and HAT-South; e.g., Bakos et al. 2004, 2007, 2013; Hartman et al. 2011; Zhou et al. 2014; and the multi-purpose CoRoT space observatory; e.g., Baglin et al. 2006; Alonso et al. 2008; Barge et al. 2008; Deleuil et al. 2008; Léger et al. 2009) were able to monitor the brightness of thousands of stars at once, continually—playing a numbers game to ensure large numbers of detections. The ultimate expression of that philosophy came with the launch of Kepler, in 2009. The Kepler spacecraft was a dedicated exoplanet finding machine—with a field of view that allowed it to continually monitor more than 150,000 stars, 24 hr per day, for four years, during which time it achieved a duty cycle in excess of 90% for many of the target stars (e.g., Garcia et al. 2014). In that manner, Kepler soon became by far the most successful exoplanet survey to date—finding some 2351 confirmed planets, and a further 2418 candidate planets that still await follow-up⁷¹ (e.g., Borucki et al. 2011; Batalha et al. 2013; Mullally et al. 2015; Coughlin et al. 2016; Morton et al. 2016).

While Kepler was a huge success, it also revealed a shortcoming of ground-based exoplanetary science work. Simply put, there were too many candidate planets to follow-up, and too few ground-based facilities to perform that work. Most radial velocity instruments are located on shared-time telescopes, meaning that researchers have to compete for time to make their observations. This, combined with the time taken to perform those observations, and the fact that a given facility can only perform radial velocity observations of a single star at a time, represents a significant bottleneck to the exoplanet detection process—and one that will only get worse as the TESS mission continues to deliver new candidate planets for follow-up work (e.g., Ricker et al. 2015; Sullivan et al. 2015; Barclay et al. 2018; Stassun et al. 2018). To address the overwhelming wealth of exoplanet candidates that will be delivered by TESS in the coming years, numerous radial velocity facilities are being employed to perform follow-up observations (e.g., Nielsen et al. 2019; Quinn et al. 2019; Vanderburg et al. 2019).

The AAPS, operated from 1997 until 2014 using the Anglo-Australia Telescope, built up a 17 yr long database of RV observations for more than two hundred Sunlike stars (e.g., Wittenmyer et al. 2016b; 2017a, Kane et al. 2019a). The Keck telescopes have also played a significant role in radial velocity searches, such as the use of HIRES by such teams as the California Planet Search, including exoplanet occurrence rates (Howard & Fulton 2016). Other facilities, such as the MINiature Exoplanet Radial Velocity Array (Minerva), and Minerva Australis (based in Australia), use off-the-shelf telescopes to reduce the cost of a dedicated facility (e.g., Swift et al. 2015; Addison et al. 2019), and have already delivered some key science results (e.g., Rodrigeuz et al. 2017; Wilson et al. 2019; Addison et al. 2020). Numerous other precision radial velocity instruments operated in conjunction with the European Southern Observatory include CORALIE (Queloz et al. 2000), HARPS (Pepe et al. 2000), and ESPRESSO (Pepe et al. 2014). Other spectrographs, such as the NEID instrument, have been developed to optimize performance in the infra-red in order to provide improved radial velocities for low-mass stars (Halverson et al. 2016). This list of radial velocity instruments is not intended to be exhaustive, and for a more detailed discussion of the future of Radial Velocity facilities, we direct the interested reader to the recent review by Fischer et al. (2016).

While the Exoplanet Era has been dominated, to date, by the radial velocity and transit methods of exoplanet detection, it is likely that, in the future, astrometric and direct imaging observations will begin to yield increasing number of exoplanet discoveries, and play an important role in their characterization. To date, direct imaging studies have been relatively limited, primarily focusing on young (and therefore hot) planets orbiting nearby stars (e.g., HR 8799; Marois et al. 2008, 2010). With recent developments, however, the number of potential targets for such imaging has increased (e.g., Kane et al. 2018). In addition, direct imaging is now finding a growing role in the confirmation of the planetary nature of objects whose existence has been inferred on the basis of observed long-period radial velocity trends (e.g., Kane et al. 2019a). In the future, this complementarity between different observing techniques will become ever more important—and astrometric observations by the Gaia spacecraft are likely to contribute large numbers of long-period exoplanets that would otherwise be particularly hard to detect with traditional methods (e.g., Perryman et al. 2014; Gaia Collaboration et al. 2016).

Another method that has long offered the promise of being able to detect significant numbers of exoplanets, and to sample a population of planets that are hard to find using other techniques, is gravitational microlensing (e.g., Gaudi 2012). Microlensing has numerous advantages over other methods, including the observational requirement of moderate aperture size telescopes and sensitivity to relatively low masses. Disadvantages include the non-repeatability of the observations and degeneracy in modeling the available data (e.g., Dominik 2009). To date, microlensing observations have discovered almost 100 exoplanets (NASA Exoplanet Archive; Akeson et al. 2013), and it is anticipated that the launch of the Wide Field Infrared Survey Telescope (WFIRST) satellite, in 2025, will lead to a new era in microlensing observations, with the potential that the spacecraft could yield hundreds, if not thousands, of new discoveries (e.g., Yee 2013; Penny et al. 2019).

Planned exoplanet missions for the future span a variety of science goals regarding both detection and characterization of exoplanets. The legacy of space-based transit detection will be continued by such missions as PLAnetary Transits and Oscillations of stars (Rauer et al. 2014), with smaller space-based telescopes like CHaracterizing ExoPlanet Satellite performing follow-up observations of known exoplanetary systems (Broeg et al. 2014). Other missions, such as Atmospheric Remote-sensing Infrared Exoplanet Large-survey (Encrenaz et al. 2018; Tinetti et al. 2018) and JWST (Boccaletti et al. 2005; Seager et al. 2009), will focus their observational efforts on detecting and characterizing the atmospheres of transiting planets via the method of transmission spectroscopy (e.g., Seager & Sasselov 2000; Kempton et al. 2018). Beyond these missions lies a pathway toward direct imaging of exoplanets and the potential for direct spectra of exoplanet atmospheres. In the near-term, the WFIRST mission will utilize a coronagraph that will enable direct imaging of giant planets at wide separations from their host stars (Kane et al. 2018; Lacy et al. 2019). Further afield, the design of the Habitable Exoplanet Imaging Mission (HabEx) is underway as a means to achieve the dream of finally being able to directly image and characterize the atmospheres of terrestrial planets, particularly those that lie in the Habitable Zone of their host stars (Kopparapu et al. 2018; Wang et al. 2018b; Kawashima & Rugheimer 2019). As both the scientific and funding landscape is a constantly evolving environment, the precise direction of future mission design and deployment will depend heavily on the discoveries that are made over the intervening years.